Journal Pre-proofs

Advances in intelligent vehicular health monitoring and fault diagnosis: Tech‐

niques, technologies, and future directions

Md Naeem Hossain, Md Mustafizur Rahman, Devarajan Ramasamy

PII: S0263-2241(25)00977-7
DOI: https://doi.org/10.1016/j.measurement.2025.117618
Reference: MEASUR 117618

To appear in: Measurement

Received Date: 28 June 2024

Revised Date: 26 March 2025
Accepted Date: 16 April 2025

Please cite this article as: M.N. Hossain, M.M. Rahman, D. Ramasamy, Advances in intelligent vehicular health
monitoring and fault diagnosis: Techniques, technologies, and future directions, Measurement (2025), doi:
https://doi.org/10.1016/j.measurement.2025.117618

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. Please note that, during the production process, errors
may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar
technologies.
Advances in intelligent vehicular health monitoring and fault diagnosis:
Techniques, technologies, and future directions

Md Naeem Hossaina, Md Mustafizur Rahmana,b*, Devarajan Ramasamya

aFaculty of Mechanical and Automotive Engineering Technology, Universiti Malaysia Pahang

Al-Sultan Abdullah, 26600 Pekan, Pahang, Malaysia

bAutomotive Engineering Centre, Universiti Malaysia Pahang Al-Sultan Abdullah, 26600

Pekan, Pahang, Malaysia

*Corresponding author email: [email protected]

Abstract
Despite numerous approaches tailored for specific components and vehicle types, the
automotive industry lacks a unified framework that integrates diverse methodologies while
addressing scalability, adaptability, and efficiency across conventional (CVs), electric (EVs),
and autonomous vehicles (AVs). This study provides a taxonomic classification of vehicle
health monitoring (VHM) and fault diagnosis techniques, highlighting the transformative role
of advanced sensors, IoT integration, artificial intelligence (AI), and multi-sensor fusion.
Through comparative evaluation of model-based, signal-based, and data-driven approaches,
we identify their strengths, limitations, and optimal applications in different vehicular contexts
supported by quantitative performance metrics. Key findings demonstrate that AI-driven fault
diagnosis in CVs achieves up to 95% accuracy in early mechanical failure detection, reducing
maintenance costs and preventing catastrophic failures. Advanced battery monitoring
techniques for EVs improve energy efficiency (15-20%) and extend battery lifespan (10-15%),
addressing critical range anxiety and sustainability concerns. In AVs, sensor fusion and AI-
based prediction achieve 99% reliability in real-time decision-making, enhancing operational
safety and passenger confidence. We propose a novel conceptual framework integrating
sensors, IoT, AI, and big data analytics to enhance VHM capabilities across all vehicle types.
This framework addresses challenges in heterogeneous data integration, efficiency, and real-
time processing in dynamic automotive environments. Future research addresses cybersecurity
vulnerabilities, optimising AI models for edge computing deployments, developing AI
techniques for transparent diagnostics, and leveraging vehicle-to-vehicle communication
protocols to improve fault detection accuracy. This study contributes to the reactive to
predictive maintenance strategies, potentially transforming maintenance practices across the
transportation sector while enhancing vehicle reliability, safety, and sustainability.

Keywords: Vehicle health monitoring, Fault diagnosis, Electric vehicles, Autonomous

vehicles, Sensor technology, Internet of Things, Artificial intelligence

1. Introduction

The enhancement of the global economy and living standards of the population enables a
staggering growth of the automotive industry and, thereby, the number of vehicles on the road.
The worldwide vehicle numbers are expected to double in the next 20 years, according to the
estimation of the European Commission [1]. The industry continues to evolve, and new
technologies have made vehicles more efficient, reliable, and safe, paving the way for more
passengers. The global automotive sector is projected to reach a valuation of USD 9 trillion by
2030 [2]. One of the key issues related to the safety and performance of automobiles is vehicle
health monitoring (VHM) and fault diagnosis systems. These intelligent and advanced systems
are critical in identifying potential faults before failure occurs, thereby playing an important
role in ensuring the safety and reliability of a vehicle on the road [3]. VHM and fault

1
identification have now become an integral part of every modern technological innovation,
from simple conventional vehicles (CVs) and electric vehicles (EVs) to still futuristic forward-
looking autonomous vehicles (AVs). The vehicle inevitably experiences failures because of
certain influencing factors including, but not limited to, external driving environment, vehicle
design structure, and aging, as well as functional anomalies in a rotating element such as metal
fatigue failure, pressure failure, temperature failure, wear and tear failure, which, materially,
can affect the driving safety of people’s lives and goods during transportation [4]. The main
goal of the VHM is to make sure the vehicle is running smoothly and dependably. Fault
diagnosis, on the other hand, involves pinpointing and locating any issues or problems within
the different subsystems of the vehicle [5]. Modern automobile systems, due to their increasing
complexity, require more advanced fault diagnosis methods than traditional model-based and
signal-based approaches, which rely on static thresholds and simple decision rules. Traditional
methods do not offer enough flexibility and accuracy to discover and prevent future faults
during live operations especially when detecting issues with EVs and AVs [6,7]. Vehicles
equipped with these outdated fault detection systems fail to identify faults accurately and in a
timely manner, leading to reduced safety and subpar performance. Moreover, the lack of
thorough diagnostic and monitoring solutions creates major financial and operational problems.
Research findings show that using poorly designed VHM techniques result in maintenance
expenses increasing by 40% [8]. Decreased failure cycles of automotive components lead to
unanticipated component malfunctions that generate unexpected downtime which interferes
with transport operations [9]. These issues have led to the development and widespread
penetration of state-of-the-art advanced intelligent technologies in VHM and fault diagnosis
techniques to overcome these limitations [10]. One example is the ability of artificial
intelligence (AI) algorithms, in particular machine learning and deep learning techniques, to
process massive amounts of data from sensors and different sources to detect patterns,
anomalies, and correlations that are indicative of faults or impending failures [11,12].
Moreover, with the advent of the Internet of Things (IoT), the data collected from these systems
are being sent continuously for analysis, adding to such intelligent networks’ dynamism.
The recent research development in the field of VHM and fault diagnosis reveals innovative
methods and technologies to improve the performance, reliability, and safety of vehicles
[13,14], as shown in Fig. 1. Choudhary et al. [15] developed an advanced review to improve
the efficiency of various components of electric vehicles depending on state-of-the-art
condition monitoring and fault diagnosis strategies. The work of Zhang et al. [16] provided a
comprehensive review of the advanced data-driven feature extraction algorithms and AI
models for electric powertrain condition monitoring and incipient fault prognostics. Based on
this rapidly evolving field, this review will provide a resource to guide the selection of
strategies to use and introduce novel AI algorithms and their relevant applications. Furthermore,
Abdeltwab et al. [17] summarised techniques for engine fault diagnosis through vibration
analysis, such as Fourier transform, higher-order statistics, Wigner-Ville distribution, and
wavelet transform, and showed the analysis as extremely helpful for feature extraction and
fault detection. While the above studies provide valuable insights into specific components,
they often lack a broader perspective, focusing narrowly on individual subsystems or vehicle
categories. These fragmented approaches fail to address the interconnected and
multidisciplinary nature of modern vehicular systems, which increasingly rely on integrated
solutions spanning mechanical, electrical, and computational domains. A critical gap exists in
the literature, as existing reviews predominantly concentrate on isolated aspects of VHM and
fault diagnosis, such as engine performance or battery management, without offering a unified
framework that encompasses the diverse and evolving landscape of CVs, EVs, and AVs.
This article distinguishes itself by providing a comprehensive and holistic review of the
most up-to-date methods and tools for VHM and fault diagnosis across all vehicle types. By

2
synthesising advancements in sensor technology, AI, IoT, and data analytics, this review
underscores the transformative potential of intelligent systems in enhancing vehicle reliability,
safety, and sustainability. Moreover, it highlights the necessity of a multidisciplinary approach,
integrating expertise from engineering, computer science, cybersecurity, and data science to
address the complex challenges posed by modern vehicular systems. By staying up-to-date on
the latest advances in these fields, the automotive industry can continue to drive innovation
and provide consumers with reliable and sustainable vehicles.

Fig. 1. near here.

In actuality, the following are included in this article:

i) A detailed taxonomic classification of VHM and fault diagnosis tailored to CVs, EVs,
and AVs, illustrated in Fig. 2, serves as a foundational framework for understanding
and improving existing methods.
ii) Recent developments in VHM and fault diagnosis involve emerging technologies such
as sensor and sensor fusion, AI, IoT, etc, to aid in fault detection, diagnosis, and
prediction of faults in the equipment.
iii) Propose a conceptual framework for VHM with fault detection, designed to address the
complexities of modern vehicular systems.

Fig. 2. near here.

The remainder of this paper is organised as follows. Section 2 presents a comprehensive

analysis of VHM and fault diagnosis systems for CVs, EVs, and AVs. Section 3 explores the
application of emerging technologies in VHM and fault diagnosis systems, while Section 4
outlines the techniques employed in this domain. Section 5 critically analyses the comparison
and trade-off between various VHM techniques. Section 6 proposes a conceptual framework
for VHM and fault diagnosis tailored to modern and connected vehicles. Section 7 discusses
the challenges and potential avenues for future research. Finally, Section 8 concludes the
review by summarising key findings and insights.

2. VHM and fault diagnosis methods

VHM and fault diagnosis systems ensure optimal safety and efficiency in operating various
vehicle types, encompassing conventional internal combustion engine vehicles, electric
vehicles, and autonomous vehicles. In recent years, significant technological advancements
have catalysed the development of sophisticated diagnostic tools and systems capable of
rapidly identifying and rectifying issues that adversely impact vehicle performance [18].
However, these systems’ specific components and requirements can exhibit variations
contingent upon the vehicle type.

2.1. Conventional vehicle

Recent advancements in vehicle health monitoring have predominantly focused on

leveraging artificial intelligence and data-driven techniques to enhance the accuracy and
reliability of fault detection across various vehicle subsystems. As depicted in Fig. 3, these
methods have shown significant improvements in classification accuracy and the ability to
perform real-time monitoring, thereby contributing to developing more robust and efficient
fault diagnosis systems.

3
 For monitoring engine performance, model-based techniques, machine learning (ML), and
deep learning (DL) approaches provide highly accurate results but can be computationally
expensive. One challenge with model-based techniques is the intricate process of designing
the observers or monitoring components. Additionally, these techniques can struggle when
dealing with datasets that are unbalanced or skewed, where some conditions or scenarios are
overrepresented compared to others [19]. In this context, ML and DL approaches such as
decision tree (DT), support vector machine (SVM), K-nearest neighbour (KNN), random forest
(RF), artificial neural network (ANN), and multilayer perceptron (MLP) methods have been
used by exploiting their capability of processing high-dimensional datasets and discovering
intricate patterns of data [20–22]. Additionally, noninvasive methods and vibration analysis
provide cost-effective and real-time solutions, but the accuracy of vibration analysis can be
affected by noise and signal processing complexity [17,23]. Moreover, hybrid approaches
combine multiple techniques to improve fault diagnosis accuracy and efficiency. For example,
Du et al. [24] proposed combining the analytic hierarchy process and neural networks to
enhance the accuracy of vehicle engine fault diagnosis by determining the weight values of
fault characteristics and using them as inputs for neural network classification.
Vibration monitoring is one of the most widely used methods for gearbox fault diagnosis.
Historical methods are based on laboratory conditions, and recent trends have focused on
studying real-world circumstances. Some improvements in classification accuracy have been
obtained by driving the vehicles on the road and carrying out time domain and frequency
domain analysis on normalised data. In the time domain, a common approach to analysing
vibration signals is calculating the root mean square (RMS) value, which quantifies signal
energy and helps detect anomalies [25]. The RMS of a vibration signal 𝑥(𝑡) over a time period
𝑇 is given by:
𝑇
1
𝑅𝑀𝑆 = 𝑥2(𝑡)𝑑𝑡
𝑇 0
Fig. 3. near here.

An increase in RMS amplitude often indicates abnormal vibration levels due to wear,
misalignment, or damage in gearbox components. For frequency-domain analysis, the Fourier
Transform is widely used to convert vibration signals from the time domain to the frequency
domain, allowing for the identification of fault-related frequency components [26]. The Fourier
Transform of a signal 𝑥(𝑡) is given by:
∞
𝑋(𝑓) = 𝑥(𝑡)𝑒―𝑗2𝜋𝑓𝑡𝑑𝑡
―∞
where 𝑋(𝑓) represents the frequency components of the signal, 𝑓 is the frequency, and 𝑗 is
the imaginary unit. By analysing the frequency spectrum, faults such as gear tooth damage or
imbalance can be detected by identifying characteristic fault frequencies. A meshless technique
to enable online condition-based monitoring has been reported, as this permits real-time signal
processing from accelerometers, leading to quicker diagnosis [27].
Wang et al. [28] proposed a clustering-based approach with Fisher discriminant analysis for
residual analysis of suspension springs. The method is model-free, which does not need to be
known as a type of suspension model and pattern to failure, and is suitable for online
monitoring. Also, observer-based techniques have been used to observe active suspension
control systems, considering vehicle nonlinearities. The system can be configured with fault
detection observers of heterogeneous subsystems to detect faults by analysing observer
residuals [29].
Raveendran and coworkers. [6] have proposed a learning-based fault identification scheme
to detect levels of brake faults that used RF and achieved a 92% classification accuracy. This

4
algorithm outperforms other methods such as Naïve Bayes, KNN, SVM, and DT. Moreover,
DT and RF have been applied to air brake systems in heavy commercial vehicles, achieving a
prediction accuracy of 94.47% [30].
For tyre monitoring, observer-based techniques and learning-based approaches offer high
accuracy and adaptability [31]. In contrast, actuator-redundancy-based methods and soft
computing approaches provide robust solutions for real-time fault detection [32]. Indirect tyre
pressure monitoring, model-based, and signal-based methods present cost-effective and
reliable alternatives for practical applications [33].

2.2. Electric vehicle

Electric vehicles have become increasingly prevalent, necessitating advanced monitoring

and fault diagnosis techniques to ensure enhanced performance, safety, and environmental
sustainability. Research has focused on leveraging model-based approaches, signal processing
methods, artificial intelligence, and machine learning to improve fault detection and diagnosis
in various EV components, including motors, batteries, and powertrain systems (Fig. 4). The
inadequate precision of the models employed for motors and batteries, particularly concerning
their lifespan estimations, imposes constraints and diminishes the robustness and accuracy of
fault detection techniques. Moreover, measurement noise renders signal-based methods
sluggish and unsuitable for detecting early-stage faults [34]. However, data-driven approaches
have shown promising results, particularly for battery management systems. For example,
Schmid and coworkers [35] developed a novel method based on cross-cell monitoring that uses
principal component analysis (PCA) to detect and localise faults in battery cells. This technique
is robust against sensor noise and dynamic load profiles, making it practical for early
fault detection. Besides, as shown in a study by Yao et al. [36], a wavelet-neural network
approach to fault diagnosis of lithium-ion batteries proposes to remove the noise signals in the
classification process through the wavelet transform, and replaces an unreasonable single-
parameter classification with multidimensional parameters. In an analogue approach for
lithium-ion batteries, a hybrid signal-based technique performs manifold learning along with
cluster-based outlier detection for abnormal signal feature determination [7]. Kaplan et al. [37]
raised a deep learning approach using long short-term memory (LSTM) networks to diagnose
electric vehicle electromechanical conversion chain faults. The developed method is
successfully simulated and practically tested for accurate fault detection based on various
sensor data.
Furthermore, hybrid approaches that combine signal-based analysis with AI techniques have
been developed to address the complexities of EV fault diagnosis. For example, an adaptive
fault diagnosis method integrates AI with signal processing to create an onboard health
monitoring system for EV propulsion systems [38]. It uses a mixed clustering technique to
prevent future failures through predictive maintenance planning. However, challenges
remain, including misdiagnosis due to similar fault signatures, complexity from multiple
influencing factors, and the inherent fuzziness or randomness of fault causes. Addressing these
issues is essential for improving the accuracy and reliability of fault diagnosis in EVs,
particularly in achieving real-time fault detection and reducing false alarms.

Fig. 4. near here.

2.3. Autonomous vehicle

Autonomous vehicles represent a swiftly evolving technology with the capacity to transform
the transportation industry. Various fault diagnosis methods have been used to identify

5
anomalies in AVs, which range from traditional model-based methods to advanced machine
learning and deep learning approaches. Among the several ML algorithms, the researchers
preferred SVM to address the issue of unbalanced datasets by incorporating linear discriminant
analysis and grey wolf optimiser in steering actuator fault diagnosis [39] and multi-sensor
systems fault identification [40]. Another research reviewed the transition from traditional to
DL methods, emphasising the effectiveness of deep neural networks, which accurately
identified faults in dynamic models [41]. These approaches highlight the potential of DL to
handle large datasets and complex fault patterns effectively. Moreover, Park et al. [42]
combined model-predictive control with sliding mode observers approaches for fault detection
and reconstruction in longitudinal control of AVs. The approach enforces invariance of
longitudinal information as it avoids false positives in radar and acceleration sensors, tracks
and validates faults based on the reconstructed relative acceleration, and compares them with
their stored prediction. Bhagavathi et al. [43] suggested a digital twin-driven technique with an
adaptive extended Kalman filter to infer the size of faults with direct parameter estimation gain
calculation over sensors. Fang et al. [44] also introduced a hybrid data analysis scheme that
uses fuzzy proportional-integral-derivative (PID) control for AV fault diagnosis. This
framework integrates sensor monitoring, anomaly detection, and actuator fault diagnosis using
the discrete wavelet transform, an extreme learning machine autoencoder, and neural networks.
Integrating fuzzy PID control helps accurately locate faults from a control perspective.

3. Emerging technologies

3.1. Sensor technology

Sensors are electronic devices that detect or measure physical attributes, such as pressure or
acceleration, and convert them into electrical signals or other readable outputs that can be used
for control systems [45]. The automotive sensors must satisfy challenging tradeoffs such as
precision, durability, ease of production, compatibility, and affordability. In the modern
automobile industry, sensors play pivotal roles in the accurate condition monitoring of various
vehicle components. Collecting data from multiple sensors placed within the vehicle makes it
possible to identify potential issues before they lead to a breakdown and diagnose faults when
they do occur. Fig. 5 depicts a summary of the sensors used in various types of modern vehicles.

Fig. 5. near here.

3.1.1. Conventional vehicle

Various aspects of standard sensors used in conventional vehicle health monitoring and fault
diagnosis have been presented in Table 1. These sensors detect abnormal pressure, temperature,
vibration, etc., of the vehicle’s engine, brake, suspension, and other components. While sensors
share the common goal of irregularities, their methodologies and focus areas are significantly
diverse. Therefore, integrating data from multiple sensors and employing advanced diagnostic
algorithms is necessary to improve the accuracy and reliability of vehicle fault diagnosis
systems. Additionally, contextual understanding and human intervention remain essential for
interpreting sensor data accurately and effectively identifying and resolving vehicle faults.
Despite their utility, sensors can present contradictory issues that complicate fault diagnosis.
False positives and negatives are common challenges. External factors such as environmental
conditions or sensor degradation can lead to erroneous readings, potentially misguiding
diagnosis efforts. Additionally, sensors may lack specificity, making it challenging to pinpoint
the root cause of a fault accurately.

6
3.1.2. Electric vehicle

Electric vehicles should be introduced to address all current vehicle problems and safeguard
the environment by installing sensors throughout the body of the EVs. Due to the internet
connectivity of the EVs, a web-based monitoring system dubbed ‘Things Speak’ has been
deployed to stay updated on the vehicle’s health in real time [46]. Sensors can help monitor
different EV systems and components, which can help to give warnings ahead of time if any
issue is going to take place so that fewer preventive measures and repairs can be done. Thus, it
will help increase the efficiency and performance of EVs. Table 2 shows the characteristics of
the sensors of the EVs. Different sensing techniques provide precise data on temperature,
voltage, current, and performance metrics parameters, leading to better efficiency and lifespan
of components. However, it can be easily damaged by digital complexity, electromagnetic
interference, and heat dissipation over time, sometimes causing the sensor's failure or reading
error.

3.1.3. Autonomous vehicle

Autonomous vehicles are very sophisticated and high-automatic modes of vehicles, so

advanced sensor technology is needed for health monitoring and fault diagnosis, which is
complex, multifaceted, and full of challenges. By incorporating data from multiple sensors,
AVs can track their own environments, identify areas of concern, and respond appropriately to
ensure that the system operates safely and reliably. Table 3 describes a few other types of
sensors with the parameters of the respective sensors to facilitate the smooth decision-making
of AVs. Together, they work to get a full view of the environment - an essential element for
safe navigation. Lidar and cameras tend to offer a higher resolution and intricately detailed
mapping of the environment, whereas radar and ultrasonic sensors are the most reliable to use
under adverse weather conditions and are more cost-effective. However, detection and
perception of AV faults face difficulties in sensor calibration and synchronisation and coping
with erratic communication errors. Consequently, by the integration and exploitation of the
advantages of the different sensors, the complete system is not only able to perform better, it
is also capable of operating in more conclusive complies as well.

7
Table 1
Overview of standard sensors used in conventional vehicles.
Temperature Monitor engine coolant temperature Silicon-integrated circuit chips, Crucial for preventing R
sensors to maintain optimal engine thermistors, resistive engine overheating, in
temperature temperature detectors enabling efficient re
operation m
te
op
Pressure Monitor brake fluid, air conditioner Inductive, capacitive, Essential for optimal S
sensors compressors, chassis, adaptive piezoelectric, potentiometric performance and fault co
suspension hydraulics, tyres, inlet detection in pressure- si
manifold, oil tank, fuel tank, fuel sensitive systems re
injection, and ambient atmosphere. co
Gas sensors Measure oxygen concentration in Oxygen concentration sensors Vital for maintaining L
exhaust gas; regulate harmful (lambda sensors) ideal air-fuel ratio, sh
emissions; monitor carbon reducing emissions, hi
monoxide, nitrogen oxide, and and ensuring in
hydrogen gas regulatory compliance pe
Position Detect abnormal behaviour in Resistance-based, capacitive, Early detection of S
sensors suspension system, steering system, optical, Hall Effect-based position-related faults, fi
brake system, throttle system, and magnetic, anisotropic enabling preventive li
transmission system magnetoresistive, giant maintenance and in
magnetoresistive, and improved safety at
magnetostrictive linear er
Speed Anti-lock braking systems, traction Measure wheel speed, engine Critical for vehicle H
sensors control systems, cruise control crankshaft and camshaft speed, stability, optimal sp
systems, engine management and transmission output shaft system performance, ef
systems, transmission control speed and driver assistance sp
systems features

8
Table 1 (Continued).
Sensor type Applications Sensing techniques Key benefits P
Mass airflow Measure air volume entering Transmit data to an engine Improved fuel economy, reduced S
sensors the engine to optimise air- control unit for fuel emissions, and enhanced engine c
fuel ratio for engine calculation performance d
efficiency t
Torque Monitor engine torque, Wireless torque Early detection of engine and T
sensors detect vibrations and measurement systems, drivetrain issues, fuel efficiency m
fluctuations, monitor fuel integrated with other sensors monitoring, and comprehensive c
efficiency, and engine like engine speed and load performance analysis c
performance analysis

Table 2
Overview of sensors used in electric vehicles.
Sensor type Applications Sensing techniques Key benefits Potential li
Battery Monitor battery condition, Shunt resistors, analog Early detection of battery Computatio
monitoring voltage swings, overcurrent, components, voltage issues, accurate SOC integration
Sensors short/open circuits, battery measurement chips, and estimation, improved energy den
aging, and state of charge NTC thermistors for battery life, and safety compatibili
(SOC) temperature readings
Motor Monitor electric motor Speed, temperature, torque, Optimised motor Heat dissip
sensors performance, detect and position sensors for efficiency, early fault interferenc
overheating, bearing wear, rotor speed, direction, and detection, improved measureme
misalignment position performance, and
reliability
Charging Monitor charging system, Temperature sensors at Safe and efficient fast Accurately
sensors cable, connectors, detect connection points, thermal charging, real-time and expecte
overheating and electrical management system, and monitoring, and with wirele
defects battery pack prevention of heat inefficiency
damage plug-in cha
Inverter Monitor inverter Temperature, current, Early detection of Reduce mo
sensors performance, detect voltage, pressure, and inverter issues, improved efficiency,
overheating, high voltage, vibration sensors efficiency and lifespan, noise levels
and overcurrent and enhanced and size of
performance

9
Table 2 (Continued).
Sensor type Applications Sensing techniques Key benefits Potent
Transmission Monitor transmission Position sensors for Improved transmission Adapt
sensors dynamics, controllability, precise rotor positioning, efficiency, reduced fuel the ne
cost, size, and efficiency commutation, and torque consumption, dynamic torque for op
control control, and compact design noise i

Table 3
Overview of sensors used in autonomous vehicles.
Sensor type Applications Key features Key benefits Pot
LiDAR (Light Obstacle detection, Active sensor, laser beam High precision and accuracy, Lim
detection and environment mapping, emission, high-resolution 3D early detection of sensor issues wea
ranging) calibration monitoring, imaging, and long-range (damage, misalignment, acc
and data consistency capability calibration errors, and enti
checking lens/mirror degradation)
RADAR Tyre condition Radio wave emission, Accurate tyre pressure and wear Lar
(Radio monitoring, obstacle range/angle/speed monitoring, precise obstacle effe
detection and detection, and collision measurement, and detection and tracking, robust reso
ranging) avoidance short/medium/long-range performance in various inte
capabilities conditions
Camera Driver monitoring, object Passive light sensor, color and Accurate driver attention Lim
recognition, and visual texture perception, and ML monitoring, road sign/traffic low
inspection algorithms light recognition, and vehicle regu
component inspection (wear,
damage)
Ultrasonic Obstacle detection, Sound wave emission, object Precise short-range obstacle Vul
automatic braking distance measurement, and detection, suitable for parking lim
short-range capability assistance, and automatic park
braking systems unr
adv

10
3.2. Multi-sensor fusion

Multi-sensor fusion is a technique that involves integrating data from multiple sensors to
collect information about the vehicle’s status, such as motion, vibration, temperature, engine
performance, and so on [110,111]. This enables higher precision as well as reliability in the
health monitoring system of the vehicle, where data can be collected from various sensors even
if one of the sensors has failed or is ineffective due to external factors. The algorithm processes
this data to form a picture from the values received from each sensor and detect deviations
from the norm. An essential benefit of multi-sensor fusion is its capability to find faults that
might be unseen by a single sensor. For example, a single fault in the engine might not be
captured using a single accelerometer sensor. However, when combined with sensors in the
engine, this is an excellent way to provide another side of the vehicle’s view that other systems
may not know [112].

Fig. 6. near here.

Based on three levels of abstraction (Fig. 6), information integration mainly includes data
fusion, feature fusion, and decision fusion. The low level of data fusion is the fusion of directly
deployed sensor data from the surrounding environment in a system. It is the foundation of
feature and inference extraction. The advantage of using this model is that it has no data loss,
contributing to higher precision [113]. Gultekin et al. [114] constructed the data fusion method
based on a convolutional neural network (CNN) with the help of the short-term Fourier
transform for the fault diagnosis system of an automated transfer vehicle. It indicates the
potential for practical use of this approach, which is doing better in accuracy than using a single
or dual-sensor approach. In feature fusion, initially, the features are extracted from each sensor
signal. Subsequently, the fusion process is carried out by integrating the extracted features to
form a more informative and discriminative feature set. A multi-sensor feature fusion approach
was implemented by Kumar and coworkers [115] for fault identification and detection of
gearbox exploiting discrete wavelet features by acquiring data from multiple sensor types
(vibration accelerometer, microphone, and acoustic emission sensors) under various operating
conditions (different loading conditions and rpm variations). Finally, the process of decision
fusion necessitates a significant degree of integration. The decision-making process involves
the initial processing of sensor data, followed by local decision-making, culminating in fusion
at the fusion centre. The decision fusion method aims to counter a particular objective decision,
and its outcomes directly impact decision-making precision. Yue et al. [116] demonstrated the
efficacy of their proposed online monitoring strategy, which employed a decision fusion
technique, by achieving a remarkable 95.4% improvement in fuel cells’ durability in hybrid
electric vehicles.
To summarise, data fusion and feature fusion focus on integrating raw data and extracted
features, respectively, while decision fusion emphasises the combination of model outputs to
achieve superior diagnostic performance. Moreover, data fusion methods offer flexibility,
feature fusion methods are accurate, and decision fusion methods ensure reliability. The choice
of fusion technique should be based on the specific requirements of the vehicle fault diagnostic
system, considering factors such as the complexity of the data, the need for accuracy, and the
importance of reliable decision-making.

3.3. Internet of Things

The Internet of Things is about the connection of objects and devices to a network (the
Internet) that can deliver data, which is the job of sensors in the environment [117]. As shown

11
in Fig. 7, a typical IoT architecture can be divided into four main layers [118]. The sensing
layer comprises sensors and actuators, which are built into the automotive components in a
manner that collects data related to various parameters like temperature, pressure, vibration,
and a few other critical parameters. Toward this end, the networking layer enables forwarding
the sensed data toward specific control units using various technologies such as WiFi, Zigbee,
Bluetooth, and 4G/5G cellular networks [119]. The data processing layer encompasses the
software and hardware components responsible for processing and integrating the
heterogeneous data streams from multiple sources within the vehicle’s IoT ecosystem [120].
This layer employs various advanced technologies and tools, including robust data
management systems, sophisticated analytics platforms, and cutting-edge machine learning
algorithms, to effectively process and analyse vast amounts of raw data. The primary objective
of this layer is to derive valuable insights and actionable intelligence from the processed data,
which can subsequently inform decision-making processes and enable proactive maintenance
strategies. Extended from the data processing layer, the application layer is the topper-level
interface of IoT architecture that provides vehicle health and operational status to the end-users
in real time. This includes a wide variety of user-accessible applications and interfaces (e.g.,
user-friendly human-computer interaction systems), mobile applications, and web portals [121].
These applications use the insights produced by the data processing layer for presenting end-
to-end and easy-to-understand information to stakeholders such as vehicle owners, fleet
managers, and maintenance personnel so that they can make informed decisions and
appropriate actions to ensure the vehicle’s continuous smooth working and safety. The highly
efficient interoperability between the two layers, provided by the high-performance IoT
support system, supports monitoring, analysing, and reporting vehicle health, modernising how
vehicles are maintained and operated [122]. When integrating more sophisticated data
processing and dashboard applications, the IoT architecture allows for a reorientation towards
predictive and proactive maintenance strategies that, with time, decrease the maintenance cost,
reduce downtime, and increase the vehicle’s reliability [123]. In addition, the application layer
enables remote diagnosis and over-the-air software updates, ensuring that vehicles remain up-
to-date with the latest software and diagnostic capabilities. It enables their supervision and
maintenance remotely from centralised sites. This feature simplifies maintenance procedures
and improves vehicle safety and reliability by detecting early warning signs, making the
necessary corrections before the vehicle becomes unsafe, and ensuring that the vehicle is
running on the latest software/firmware with security patches and performance improvements.

Fig. 7. near here.

IoT is emerging as a transformative technology, offering significant advancements in real-

time monitoring, data acquisition, and predictive maintenance of automobiles. Recently, Wang
et al. [124] demonstrated the use of IoT technologies for online monitoring of automotive
engine lubricating oil, utilising neural networks and spectral data to identify wear states and
predict potential faults with high accuracy. Moreover, for autonomous vehicles, Jeong and his
team [125] proposed an integrated self-diagnosis system leveraging IoT gateways and deep
learning to enhance fault detection efficiency, reduce system overhead, and improve safety
while reducing costs. Several methodologies have been proposed for IoT-based vehicle fault
diagnosis. One approach involves using knowledge-based fault diagnosis systems, which
enhance interoperability and provide high-level reasoning and inquiry responses to non-expert
users. These systems utilise ontologies and deductive or inductive reasoning to construct
knowledge bases, enabling effective fault diagnosis in complex IoT environments [126].
Another methodology focuses on the real-time collection and analysis of multi-variable data
using wireless sensor networks, such as Zigbee technology. These systems can achieve

12
accurate state estimation and intelligent fault prediction by employing particle filtering
algorithms, enhancing fault diagnosis systems’ diagnostic level and scope [127]. However,
integrating various technologies and communication solutions in IoT-based systems can be
challenging, and ensuring seamless communication and data exchange between different
system components is crucial for the effective functioning of fault diagnosis systems [128].

4. Intelligent techniques

4.1. Artificial intelligence and big data analytics

AI and big data analytics have become integral to VHM and fault diagnosis, as they enable
data-driven decision-making by analysing vast amounts of sensor data collected from vehicles
[129]. AI-based predictive maintenance involves real-time data acquisition from engine
components, EV batteries, gearboxes, and braking systems. By processing these data using ML
and DL models, patterns indicative of potential faults can be detected early, reducing
unexpected breakdowns and maintenance costs [130]. While condition-based monitoring
provides direct fault detection through real-time sensor analysis, statistical predictive
maintenance leverages historical vehicle data to predict failures. Big data analytics further
enhances fault diagnosis by integrating fleet-wide information, including vehicle age, mileage,
and maintenance history, allowing for more comprehensive predictive modelling. However,
challenges such as noisy sensor data, variations in operating conditions, and the need for high-
quality labelled datasets can limit the effectiveness of AI-driven maintenance strategies.
Moreover, reliance on cloud computing for data storage and processing introduces latency
issues, which may hinder real-time fault detection in highly dynamic vehicular environments
[131].

4.1.1. Machine learning

Machine learning is a subset of artificial intelligence that enables machines to learn

automatically and improve their functionality from data without being explicitly programmed.
It involves using algorithms and statistical models to allow devices to make decisions based on
input data [132]. ML comprises two crucial stages, namely training and testing. A model is
trained during the training phase based on empirical data. During the testing phase, predictions
are generated by utilising the trained model [133]. In classification-based fault diagnosis, ML
models aim to minimise a loss function, commonly cross-entropy loss, defined as:
𝑁

𝐿=― 𝑦𝑖 log(𝑦𝑖)
𝑖=1
where 𝑦𝑖 represents the actual class label, 𝑦𝑖 is the predicted probability of the class, and 𝑁
is the number of samples. This loss function ensures that the predicted probabilities align with
actual fault classifications, improving diagnostic accuracy. Three main ML types are used in
VHM and diagnostic systems: supervised learning, unsupervised learning, and reinforcement
learning [134], summarised in Table 4. Supervised learning models are effective in fault
classification due to their high accuracy in detecting and isolating faults, as demonstrated in
vehicle power transmission systems and autonomous vehicle controllers [40,135]. However,
supervised models are highly dependent on large labelled datasets, which makes them
impractical for real-world vehicle fault diagnosis due to the difficulty in obtaining annotated
fault data across different vehicle models and operational conditions. In contrast, unsupervised
learning approaches offer the advantage of automatic feature extraction and reduced human
intervention, making them suitable for handling large datasets and non-stationary sensor faults

13
in autonomous cars [136]. In clustering-based fault detection, K-Means minimise intra-cluster
variance, defined as follows:
𝑘

‖𝑥 ― 𝜇𝑖‖2
𝑖=1 𝑥∈𝐶𝑖
where 𝐶𝑖 represents the 𝑖𝑡ℎ cluster, 𝜇𝑖 is the centroid of cluster 𝐶𝑖, and 𝑥 is a data point. This
allows ML models to identify fault patterns in unlabelled datasets, making them highly useful
for predictive maintenance in modern vehicles. Yet, unsupervised methods may struggle with
feature similarity and shift-variant properties, potentially leading to misclassification.
Reinforcement learning, despite its potential to optimise real-time fault prediction and
maintenance strategies, is hindered by long training times, high computational costs, and the
challenge of defining suitable reward functions for complex vehicle dynamics [137]. The
contradiction among researchers lies in whether supervised methods should remain the gold
standard due to their proven accuracy or if unsupervised and reinforcement learning approaches,
despite their interpretability and automation challenges, can offer scalable solutions for diverse
and dynamic vehicle health monitoring environments. This underscores the need for integrated
hybrid frameworks, where supervised models improve fault classification, unsupervised
techniques enhance anomaly detection, and reinforcement learning optimises predictive
maintenance strategies.

Table 4
Comparative analysis between machine learning approaches.
ML Types Supervised Unsupervised Reinforcement Ref.
Criteria
Definition Learn from Find hidden Learn through [138–140]
labelled data patterns in rewards and
unlabelled data penalties
Components Regression and Clustering, Policy, value [141–144]
classification association function, and model
rules, and
dimensionality
reduction
Common SVM, RF, K-means Q-learning, deep Q- [39,135,145–150]
algorithms KNN, DT clustering, PCA, networks
autoencoder
Used cases Fault Anomaly Decision-making [135,145,151–
classification, detection in for controlling, 153]
failure bearings, gears, navigation, and
prediction, and and nonlinear obstacle avoidance
warning in- dynamic of autonomous
vehicle systems systems to vehicles
like identify and
transmission, predict faults
engine, and tyre
Data Large and Large and Interaction data with [138,139,154]
requirements labelled datasets unlabelled the environment
datasets
Model Moderate to Moderate High [145,148,151,152]
complexity high
Scalability Good with Highly scalable Can be complex due [151,152]
appropriate data to environmental
interactions

14
Table 4 (Continued).
ML Types Supervised Unsupervised Reinforcement Ref.
Criteria
Accuracy Generally high Varies High but dependent [155–157]
if well-trained depending on on training quality
data quality and and environment
algorithm
Computational Moderate to Moderate High [141–143]
cost high
Deployment Relatively Requires a good Complex due to the [149,150,153]
straightforward understanding of need for continuous
once trained data patterns learning

4.1.2. Artificial neural network

An artificial neural network is a machine learning model class inspired by the human brain’s
neural networks. The basic structure of an ANN consists of a network of neurons from various
layers- input, hidden, and output, which are densely connected [158]. Raw data (e.g., sensor
readings or images) enters an input layer and flows through one or more hidden layers, each
containing several neurons connected via synaptic connections. Every neuron in hidden and
output layers computes a regression of the weighted sum of inputs from previous layers of
neurons. It uses a nonlinear activation function to bring non-linearity to its output [159].
Mathematically, a neuron’s output can be expressed as:
𝑛

𝑦=𝑓 𝑤𝑖𝑥𝑖 + 𝑏
𝑖=1
where 𝑥𝑖 represents the inputs, 𝑤𝑖 are the corresponding weights, and 𝑏 is the bias term. One
of the strongest sides of ANNs is, unlike conventional models or SVMs, its capacity to learn
very complex, hierarchical features and patterns from data during the model fitting process.
The remaining part of the neural network (the optimiser) will be an iterative optimisation
process that adjusts the interconnection weights between neurons and the associated biases
such that the discrepancy between the network’s predictions and the desired outputs is minimal.
Due to this training process, the ANNs can capture complex nonlinear relationships and latent
patterns from the data, which makes them particularly good at fault diagnosis, prognostics, and
health management in complex systems such as vehicles. In addition, ANN varied from
shallow architectures with only one hidden layer to deep neural networks with multiple stacked
hidden layers. In particular, deep neural networks (DNNs) have exhibited state-of-the-art
performance in many applications thanks to their capability to learn hierarchical abstraction
features and capture complex data distributions. Their deep and multi-layer architectures
endow ANNs with strong hierarchical feature learning capabilities, so they can be applied to
analyse and learn the features of substantial vehicle data and may be used for vehicle fault
diagnosis and prognostics applications.
Another critical element of ANNs is how they learn during the training stage when the
weights of the interconnections between neurons are iteratively adjusted to minimise the error
between the predictions made by the network and the actual targets. Backpropagation is the
most popular algorithm that achieves this weight optimisation. It computes the gradient of error
concerning weights and biases, enabling updating the values of weights and biases to minimise
the error [160]. Due to the versatility of ANNs, they have been used to optimise vehicle
network systems, and backpropagation neural networks have been used to enhance the
intelligence and accuracy of fault diagnosis in vehicle controller area networks. Much research
has illustrated that ANNs can diagnose failure modes in EVs or conventional vehicles, where

15
they preserve the power of non-linear model generation by treating complex data patterns
efficiently. For instance, Xue et al. [161] have adopted advanced hydrocarbon artificial
networks to identify mechanical faults in in-wheel motors in actual electric vehicles, proving
that it can also demonstrate the system’s robustness under normal operating conditions and
influence. Similarly, Du et al. [162] proposed that the anti-jamming SVM with particle swarm
optimisation can be used together in the fault diagnosis of an automobile engine, which offers
diagnostic fault tolerance and practical application convenience without feature sample
preparation and manual feature extraction. The success of ANNs in fault diagnosis for a vehicle
is attributed to their high-order pattern recognition and generalisation abilities, which can be
achieved by learning complex nonlinear relationships through copious operation conditions,
sensor data, etc. As a result, they are ideal approaches for identifying subtle anomalies and
degradation patterns that can otherwise be challenging via conventional rule-based or model-
based techniques.

4.1.3. Deep learning

Automatic vehicle health monitoring and fault diagnosis systems have evolved as one of the
promising breakthroughs for vehicle reliability, and VHM research and development have
become an enormous concern in the automotive industry. Deep learning, a sophisticated field
of machine learning, has exhibited impressive results and gained fame in every domain. DL
architectures, such as convolutional neural networks, recurrent neural networks, long short-
term memory networks, multilayer perceptrons, deep belief networks, etc., have demonstrated
exceptional performance in automotive fault diagnosis. DL is built on artificial neural networks
with many layers of statistical units. These layers allow for the learning of complex data
representations edge-to-edge. This makes them very suitable for mastering complex and typical
problems of the modern automotive industry [163]. DL has been very successful in fault
diagnosis because it can automatically learn the most informative and discriminative features
from high-dimensional, multimodal data sources (i.e., sensor readings, imagery, and time-
series data), which is otherwise tedious, labour-intensive, and biased due to human intervention
[164]. For instance, deep symptom-based models have achieved unprecedented performance
in predicting various vehicle faults based on historical data, exceeding state-of-the-art results
in accuracy, precision, recall, and F-Score [165]. Moreover, DL models can capture complex
non-linear relationships and patterns in data, making them effective at identifying subtle
anomalies and fault signatures that are hard to detect with traditional methods. However, the
success of DL models depends on the availability of large volumes of high-quality labelled
data, which can be challenging to obtain in automotive environments. Additionally, DL models
are computationally intensive, requiring specialised hardware accelerators such as graphics
processing units for training and inference [166]. To address these challenges, techniques such
as transfer learning and data augmentation are increasingly used. Transfer learning leverages
pre-trained models to improve performance on new tasks, while data augmentation generates
synthetic data to enhance model robustness. Some of the most modern neural networks use
transfer learning to alleviate the scarcity of labelled data and expedite the training process.
Deep transfer learning methods that fuse the strengths of deep learning and transfer learning
together have significantly enhanced accuracies of fault detection and generalisation
capabilities in different datasets and domains [165].
Combining DL with other emerging technologies, such as the IoT, edge computing, and 5G
communication, offers possibilities for real-time, decentralised, and scalable implementation
solutions for fault diagnosis, resulting in proactive maintenance and improving the safety and
reliability of vehicles overall. Additionally, adversarial learning frameworks, such as deep
adversarial convolutional neural networks, have effectively addressed challenges associated

16
with limited labelled data and poor generalisation to unseen conditions, thereby significantly
enhancing the robustness and applicability of diagnostic models [167]. The ensemble deep
learning approach, which integrates multiple DL models, has also proven effective in
diagnosing complex faults in the rotor and bearing systems, demonstrating better adaptability
than single models [168]. A recent study by Luo et al. [169] demonstrated the capability of DL
in accurate damage prediction of automotive suspension systems with varying real-world
operating conditions. They introduced dual-tree complex wavelet enhanced convolutional long
short-term memory. This new neural network architecture achieved better performance for the
prediction of degradation of the suspension components than other baseline approaches. This
research stresses the robust and better performance of the adaptable DL methods toward highly
complicated fault detection and prognosis tasks for the automotive sector. Despite these
advancements, operationalising DL models in practical fault diagnosis systems remains a non-
trivial task, requiring careful consideration of data quality, model interpretability, and
computational resource constraints. Researchers are divided on whether hybrid approaches
integrating DL with traditional machine learning and physics-based models offer better
reliability than purely data-driven methods. It underscores the necessity for explainable AI
techniques and domain adaptation strategies to enhance DL’s applicability in real-world
vehicle health monitoring, ensuring a balance between accuracy, efficiency, and
interpretability.

4.1.4. Fuzzy logic

Fuzzy logic is a mathematical framework for dealing with uncertain or imprecise

information. Unlike classical (or Boolean) logic, which is based on binary (true or false) values,
fuzzy logic allows for partial truths, with values ranging from 0 (completely false) to 1 (entirely
true) [170]. Fuzzy systems have emerged as a promising technology for vehicle fault diagnosis,
offering robust solutions to the complexities and uncertainties inherent in automotive systems.
Applying fuzzy logic in fault diagnosis allows for handling imprecise and noisy data, which is
crucial for accurately identifying faults in vehicle actuators and sensors. For instance, the use
of Takagi-Sugeno fuzzy observers has been shown to effectively detect faults in steering and
torque actuators of autonomous ground vehicles, leveraging nonlinear models to reduce design
conservatism and numerical complexity, thus enhancing real-time application feasibility [171].
Additionally, fuzzy-based algorithms have been developed by Mayadevi et al. for diagnosing
drive faults in induction motor drive systems, utilising parameters such as root-mean-square
value and total harmonic distortion of stator currents to characterise and diagnose faults
accurately [172].
Fuzzy logic has become a robust VHM and diagnostics methodology. It takes rules to
interpret this complex data, which can be used to make decisions about the state of health [173].
Fuzzy logic for VHM and fault diagnosis starts with specifying a group of linguistic variables.
Those are the variables that say what the vehicle is doing or in which mode it is running, such
as the temperature of the coolant or the engine rpm, etc. Each attribute is fuzzified into an
ensemble of fuzzy sets corresponding to various degrees around that attribute [174]. For
instance, do the fuzzy partitioning of the coolant temperature variable temperature examples
as: ‘cold’, ‘warm’, and ‘hot’. It is defined by membership functions, which specify the extent
to which a particular temperature value belongs to the set. Fuzzy logic enables solving
problems with imprecise or inexact information, a desirable capability for operation in complex,
dynamic vehicle environments. A set of fuzzy rules can be designed to map the input variables
to output variables (fault diagnosis or health status) by using expert knowledge and modelling
experience. Additionally, the fusion of fuzzy logic with other intelligent systems, such as ANN,
has been suggested to improve diagnostic capability [175]. Using neural networks allows the

17
formulation of fuzzy rules to be learned and adapted from historical information, which
provides a more reliable and accurate fault diagnosis in real-time situations.

4.2. Dynamic Bayesian network

Dynamic Bayesian Networks (DBNs) use probabilistic graphical models to illustrate

complex systems that evolve. DBN extends the concept of Bayesian networks to modelling
temporal dependencies between variables, which is very suitable for modelling and predicting
the behaviours of dynamic evolving systems [176]. A DBN is a directed acyclic graph that
captures the probabilistic relationship between each node, where a node represents a random
variable and an edge connects two nodes and describes the relationship between two random
variables [177]. The joint probability distribution of a DBN over multiple time slices can be
expressed as:
𝑇

𝑃(𝑋1:𝑇, 𝑍1:𝑇) = 𝑃(𝑋1, 𝑍1) 𝑃(𝑋𝑡|𝑋𝑡―1, 𝑍𝑡) 𝑃(𝑍𝑡|𝑍𝑡―1)

𝑡=2
where 𝑋𝑡 represents the state variables at time 𝑡, 𝑍𝑡 represents the observations at time 𝑡,
and 𝑃(𝑋𝑡|𝑋𝑡―1, 𝑍𝑡) defines the transition probabilities capturing system evolution. DBNs have
been successfully applied to detect and diagnose faults in different vehicle systems. For
example, Gomes et al. [178] applied a component-based health monitoring system using DBNs
in a hierarchical way to the autonomous vehicle with a high level of missing data to detect,
diagnose, and predict faults successfully. These functions are essential in ensuring autonomous
vehicles’ safety and reliability. In another study by Xue et al. [179], DBNs were used to
diagnose electric vehicles with in-wheel motors. To adopt frequent acceleration and
deceleration, the researchers created a model that manages the unstable electrical signals
produced by this type of test and increases the reliability of fault detection. These researches
show that the DBNs are flexible with various vehicle parts and operation conditions. Another
reason is that DBNs are very useful in dealing with uncertainties and counteracting some more
complicated fault scenarios. For instance, Wang et al. [180] used a network for diesel engine
fuel injection system diagnosis to deal with the fact that fault types are closely mutual and
coupling with each other.
A remarkable case of integration like that is presented by Bodrumlu et al. [181], utilising
DBNs combined with an extended Kalman filter (EKF) for lateral dynamic system fault
recognition in a ground vehicle. EKF contributes its state estimation capabilities for dynamic
systems and compensation for non-linear phenomena, and it was the component responsible
for faulty speed measurement detection in the hybrid approach. The DBN component
augmented EKF by pinpointing precisely which sensor had a fault, taking advantage of the
temporal DBN’s modelling and uncertainty propagation preferred hybrid system achieved a
fault detection probability of more than 92% by combining the outputs of these two methods,
showing the capability of DBNs in combination with other methodologies for achieving better
fault diagnosis performance. The fusion of DBNs and other diagnostic procedures can be
expanded to integrate with techniques like particle filters, neural networks, fuzzy logic systems,
etc. In this case, some reasonable fault diagnosis solutions can be obtained based on these
combined approaches, complementing the advantages of every single approach. For example,
particle filters can accommodate non-Gaussian noise distributions and highly non-linear
systems, and neural networks are powerful tools for pattern recognition and learning.
Sometimes, a Bayes filter gets close to zero because it avoids the wrong data instead of
collecting the correct data. At the same time, fuzzy logic systems can accurately represent
uncertainties and inexact information (very common in vehicular systems).

18
5. Comparative analysis and trade-offs between fault diagnosis methods

Table 5 presents a comparative analysis of fault diagnosis techniques across conventional

vehicles, electric vehicles, and autonomous vehicles, based on accuracy, cost, and real-time
applicability. Deep learning and multi-sensor fusion techniques offer the highest accuracy,
particularly in AVs and EVs, due to their ability to process large-scale data efficiently.
However, these methods come with higher computational requirements. Model-based and
signal-based methods remain cost-effective alternatives, but they generally exhibit lower
accuracy and real-time performance. IoT, deep learning, and multi-sensor fusion techniques
provide the best real-time applicability, making them highly suitable for modern EVs and AVs,
where real-time decision-making is crucial for safety and performance. The choice of fault
diagnosis method depends on the trade-off between cost, computational feasibility, and
required accuracy for different vehicle applications, which has been presented in Table 6.
While AI-driven fault diagnosis methods enhance accuracy, they also introduce computational
challenges, particularly in resource-constrained vehicular environments. Deep learning models,
such as CNNs and DNNs, offer superior accuracy (up to 99.5%) but require high computational
power and memory, making them less suitable for embedded vehicle systems without high-
performance processors. Machine learning methods like SVMs and random forests RFs
provide a balance between accuracy and computational efficiency, making them ideal for real-
time fault detection in EVs and AVs. Bayesian networks and model-based approaches, while
computationally lightweight, may trade off accuracy due to their dependence on predefined
rules and models rather than adaptive learning. IoT-based methods leverage cloud computing
to offload computation but introduce latency and require stable connectivity. The optimal
selection of AI models depends on the specific vehicle’s computational resources, the need for
real-time processing, and the trade-offs between accuracy, latency, and power consumption.

19
Table 5
Comparison of fault diagnosis techniques for different vehicle types.
Fault diagnosis Vehicle Accuracy Cost Real-time applicability Strengt
technique type
Model-based CV Moderate Low Limited real-time applicability due to Low co
methods complex modelling requirements. interpr
EV High for well- Moderate Limited by model accuracy and results
defined computational complexity
models
AV Moderate High Computationally expensive, requiring
significant sensor fusion
Signal-based CV Moderate Low Can be real-time but prone to noise Fast
analysis interference process
EV High Moderate Requires additional filtering for real-time cost-ef
applications
AV Moderate High Less effective in dynamic conditions with
complex sensor data
Machine learning CV High Moderate Effective in real-time applications with Good
optimised models accurac
EV High Moderate Real-time feasibility depends on data adaptiv
to high availability and computational power learnin
AV Very high High Can operate in real-time but requires
significant computational resources.
Deep learning CV Very high High Effective but requires extensive labelled Highes
data for training accurac
EV Very high High Real-time applications possible with edge learns c
computing support pattern
AV Extremely Very high Highly suitable for real-time applications
high but computationally expensive
Multi-sensor CV High Moderate Improves real-time fault detection Real-ti
fusion capabilities robust,
EV Very high High Ensures accurate and real-time fault multipl
prediction with IoT integration sources
AV Extremely Very high Essential for real-time fault detection and
high self-diagnosis

20
Table 5 (Continued).
Fault diagnosis Vehicle Accuracy Cost Real-time applicability Strengt
technique type
Internet of CV Moderate Moderate Limited by connectivity issues in older Remot
Things-based vehicles monito
EV High High Effective for remote monitoring and predict
predictive maintenance mainte
AV Very high Very high Integral to AV operation, enabling real-time
fault management
Bayesian CV High Moderate Effective for probabilistic fault diagnosis Handle
networks but needs well-defined models uncerta
EV High Moderate Suitable for handling uncertainties in probab
battery and motor diagnostics reasoni
AV Very high High Essential for probabilistic decision-making
in complex AV environments

21
Table 6
Trade-off between VHM methods.
AI Model Accuracy Computational Trade-offs Recommendations Ref.
feasibility for optimisation
Deep Very high Low (high High accuracy Model [37,168
learning computational but pruning, quantisation ,182]
demands, requires computationally , and edge computing
graphic intensive; (to reduce latency
processing units/ unsuitable for and computational
tensor processing low-power load)
units systems
Machine High Moderate (runs on Good balance Optimise feature [155–
learning low-power between engineering and 157]
embedded accuracy and use automated
systems) computational feature extraction
feasibility
Bayesian Moderate Moderate Effective for Implement parallel [178,18
networks to high (efficient for probabilistic processing to 3]
small datasets but reasoning but improve
resource-intensive limited real-time computational
for large-scale applicability efficiency
applications)
IoT-based High Moderate High accuracy Optimise data [119,18
techniques (depends on but requires transmission (e.g., 4]
communication robust compress data, use
infrastructure and communication efficient protocols
edge/cloud networks like message queuing
computing) telemetry transport,
Zigbee)
Signal Moderate High Lower accuracy Combine with ML [17,23]
processing (computationally but highly models for hybrid
efficient, runs on feasible for real- approaches to
low-power time improve accuracy
systems) applications

6. Proposed framework for VHM and fault diagnosis

The vehicle health monitoring and diagnosis framework enables a preemptive vehicle
maintenance and repair strategy by continuously observing all the critical vehicle components,
detecting faulty components in real time, and triggering preventive maintenance routines and
activities before any incidents and accidents. The proposed framework shown in Fig. 8 can be
characterised as follows.
The process starts with a sensor installation on different types of vehicles, including CVs,
EVs, and AVs, to track their performance continuously. They focus on collecting sound waves
from the operation of the different parts of the vehicle [189]. An obvious choice is using high-
resolution sensors to ensure precise and accurate measurements. The raw data containing
individual sensor or sensor fusion data is transferred to IoT gateway devices (message queuing
telemetry transport, hypertext transfer protocol, etc.) through different communication
protocols, Bluetooth, WiFi, 4G, Zigbee, etc. Accompanying the sound wave signals collected
by the sensors, the data may incorporate relevant information like vehicle speed, temperature,
and pressure. The acquired signals are susceptible to background noise, such as friction and
collision of vehicular parts. Consequently, feature extraction is performed through various
filtering algorithms, including linear predictive coding, mel frequency cepstral coefficient, and

22
wavelet transform, to eliminate the background noise and obtain processed signals while
preserving the salient features [190]. The resulting filtered signals are subjected to edge or
cloud computing data processing. Edge computing processes data locally at the vehicle level
to minimise latency and ensure real-time fault analysis. Cloud computing, on the other hand,
provides large-scale storage and powerful computational resources for extensive data analysis.
Various machine learning or deep learning algorithms are employed to analyse the filtered
signals and identify vehicular anomalies. These algorithms are trained on a dataset containing
diverse signals corresponding to different fault conditions, enabling them to classify the signals
into distinct categories based on the specific fault condition [191]. Upon fault detection, the
system diagnoses the issue and its severity by analysing the signals collected by the acoustic
sensors and other relevant data collected by the vehicle’s onboard systems (e.g., temperature,
and pressure) [192]. The diagnosis is then sent to the driver, fleet manager, or other
stakeholders through an onboard vehicle-monitoring dashboard or a remote monitoring system.
The alert provides information about the fault as well as the level of severity so the owner or
service provider can take the appropriate action. Organised maintenance repair job activities
can be scheduled much earlier using fault diagnosis alerts, thus preventing incidents or
accidents. Based on an assessment of how serious the identified fault is, the system has the
capability to advise on the nature of the repair needed as well as indicate how urgently that
repair should be undertaken. In addition, it allows the acoustic sensors to collect information
and then generate reports and statistics about the vehicle’s observational health, which means
patterns and trends in the faults and issues exist. By improving vehicle performance, these
reports can increase the safety and reliability of vehicles and help vehicle manufacturers design
and maintain better vehicles.

Fig. 8. near here.

6. Prospects, future directions and solutions

The rapid advancement of vehicle technologies, particularly in the realms of connectivity,

autonomy, and electrification, has introduced unprecedented opportunities and challenges for
vehicle health monitoring and fault diagnosis. As vehicles become increasingly complex and
data-driven, the need for innovative solutions has never been more critical. Traditional
approaches to fault detection and maintenance are no longer sufficient to address the dynamic
and interconnected nature of modern vehicular systems. Instead, the integration of cutting-edge
technologies - such as AI, IoT, and advanced sensor networks - has emerged as a transformative
force in the automotive industry. However, realising their full potential requires addressing key
challenges (Fig. 9), including data integration, cybersecurity, and the development of user-
centric systems. This section critically examines the prospects and future directions for VHM
and fault diagnosis, highlighting the innovative solutions that are driving the industry toward a
smarter, safer, and more sustainable future.

Fig. 9. near here.

6.1. Data integration from multiple sources

The integration of data from multiple sources - such as onboard sensors, control units, and
external systems can significantly enhance the accuracy, reliability, and efficiency of fault
detection and diagnosis [193]. By consolidating data streams from these heterogeneous sources,
a more comprehensive and holistic view of a vehicle’s health can be achieved, enabling
proactive maintenance and reducing the likelihood of false alarms. One of the primary

23
challenges in this domain is the effective management and processing of massive, real-time
data streams. Advanced research should develop contemporary algorithms that combine shared
analytical methods with big data techniques for the cloud-based processing of standardised
data formats [194]. Multiple sensors such as vibration alongside temperature, pressure, and
speed sensors provide their output data to advanced data fusion approaches that enhance fault
scanning accuracy while reducing noise impacts. Real-time data processing needs can be
solved through a combination of cloud computing services and data processing technologies
located near the sources known as edge computing. Through cloud platforms fleets obtain data
storage space that enables them to easily perform large-scale predictive maintenance while
sharing diagnostic insights with other users. Edge computing enables instant data processing
within vehicles so faults are detected more quickly, especially in restricted communication
settings. The implementation of virtual dynamometers provides an effective method to
calculate engine torque output for performance assessment while measuring data points [195].
An AI-based system implementing signal integrity monitoring provides an advanced method
to detect complex sensor data anomalies by helping identify faults from signals while reducing
incorrect diagnostic results [196]. Continuous monitoring techniques when used with ML
algorithms enable professionals to spot fault patterns humans would normally overlook.

6.2. Cybersecurity

Modern interconnected vehicles face growing cyber threats that include unauthorised entry
to information systems and breaches of data while operating under malicious attacks. Open
vulnerabilities impact three core diagnostic data aspects including integrity, confidentiality,
and availability which create substantial safety risks for vehicles at operation [197]. The
trustworthiness of VHM systems along with secure fault diagnosis operation requires the
implementation of powerful cybersecurity measures. Upcoming research needs to concentrate
on building advanced cybersecurity systems which defend data from initial transmission to
storage until processing [198]. All frameworks protecting VHM systems should incorporate
secure communication protocols in addition to data encryption functionality along with
authentication methods and access control solutions and real-time security monitoring tools.
Data protection is achieved through encryption technology, which prevents interception and
tampering of diagnostic data, and authentication methods, which restrict access to virtual
private networks. Real-time threat detection systems alongside their ability to detect
cyberattacks before damage happens guarantee the continued reliability of health monitoring
systems [199]. Scientists focus research on creating new cybersecurity protocols which target
vehicular networks specifically. The protocols must handle existing vulnerabilities as well as
new threats which emerge on the horizon. The evaluation of VHM system robustness combined
with the development of effective countermeasures receives significant support from
ComFASE which provides communications fault and cyberattack simulation capabilities [200].
Manufacturers can detect weaknesses in their systems through these tools to establish advanced
security measures that strengthen total system resistance. Modern cybersecurity frameworks
gain significant enhancement through the installation of AI-dependent and ML-based
analytical systems which lead to better monitoring of threats. AI systems maintain the
capability to evaluate large datasets in real time in order to spot abnormal patterns that indicate
cyberattacks. Network traffic monitoring through anomaly detection algorithms together with
predictive analytics that bases its threat predictions on historical data trends demonstrates these
technology capabilities according to Yang et al. [201]. Such technologies create enhanced
capabilities to both spot and react to cyber threats promptly and effectively.

6.3. Prognostics and health management

24
 The prognostics and health management (PHM) system brings a revolutionary maintenance
strategy which relies on predicting maintenance needs through data-based models of equipment
health state and failure predictions. Implementing historical and current data enables PHM
systems to determine critical components remaining useful life (RUL) and spot potential
failures in advance [202,203]. This proactive approach not only enhances vehicle reliability
and safety but also optimises maintenance schedules, reducing operational costs and
minimising downtime. Future research in PHM should prioritise improving the accuracy and
reliability of predictive models through the integration of advanced techniques such as ML
algorithms, DBNs, and cloud-based analytics. For instance, ML models can analyse vast
amounts of sensor data to identify patterns indicative of component degradation, while DBNs
can model the temporal dependencies and uncertainties inherent in vehicle systems [178,204].
PHM systems gain strength when united with vehicle-control systems and cloud platforms
because they enable full-time health monitoring and better predictive analysis. These systems
achieve predictive capability by determining component RUL therefore practitioners can
conduct maintenance at appropriate times. PHM systems connected to AI-based health
monitoring produce insightful findings from analysis of engine performance together with
battery health and tyre conditions data to minimise maintenance expenses [196]. PHM
frameworks which integrate edge computing with IoT technologies provide substantial power
to process data in real time and make instant decisions, according to Ren et al. [205]. The
analysis of vehicle data takes place within the vehicle through edge computing to deliver fast
and immediate responses for detected anomalies. Platform networks through IoT technology
enable easy data transmission to cloud system databases where researchers conduct analyses
and perform ongoing trend assessments.

6.4. Automated fault diagnosis

Automated fault diagnosis systems use wireless sensors, and IoT connectivity and high-
definition cameras together with advanced technologies to track vehicle surroundings and
internal systems [206]. The combination of ML models enables these systems to process real-
time data which helps them detect problems automatically without human operator
involvement. Automated fault detection systems face the main obstacle of getting self-
diagnostic capabilities in AVs. A vehicle achieves self-awareness when it can evaluate its state
of health and find faults to respond with necessary corrective actions. Future investigations
should develop durable and implementable automated fault detection platforms for
diversifying and dynamic operational settings. Researchers must create algorithms which
effectively deal with real-world driving conditions’ natural variability like weather factors,
road conditions, and traffic situations [207,208]. Research should analyse how edge computing
and cloud-based analytics can improve real-time processing capability when combined with
the system for making decision-based responses. Self-learning systems represent an essential
direction because they develop the capability to enhance their diagnostic performance through
continuous learning processes. These systems utilise reinforcement learning together with
transfer learning to adapt their fault detection capabilities to new patterns as well as develop
vehicle technologies [209].

6.5. Human-machine interface

Human-machine interfaces (HMIs) provide the main method for drivers along with
maintenance personnel and other stakeholders to connect with and analyse vehicle health
information. HMIs that function effectively allow users to quickly grasp important data

25
including fault diagnoses as well as maintenance alerts and system performance metrics which
facilitates quick and well-informed decision-making [210]. Future work should target HMI
development because research opportunities exist to design better systems which users can
customise and understand in their operational context. HMI design research needs to
concentrate on usability, customisation features, information density, and context-awareness
capabilities. An interface system can show only essential data which is determined by both user
roles (e.g. driver, technician, or fleet manager) and present conditions of the drive (e.g. road
conditions or maintenance requirements). System users benefit from multimodal interfaces
when they can interact with the system through voice commands together with touchscreens
and visual displays which enhance both user experience and accessibility [211]. Voice-based
interfaces deliver real-time health system messages to drivers without requiring them to switch
their eyes from driving yet touchscreens provide thorough maintenance data for technicians at
repair times. Researchers explore vital developments to create HMIs that change their
displayed content and visual methods in line with users’ characteristics and operating situations.
Manual learning systems use ML models to discover user behaviour patterns so they can
display targeted information which reduces mental strain for better decision outcomes [212].
Expert research presents promising ways to use augmented reality (AR) technology in HMIs
for improving vehicle health data visualisation [213]. The combination of physical elements
with AR-based interfaces permits technicians to preview diagnostic information that overlays
onto the physical components thus improving their diagnosis and repair processes. The
integration of an AR headset would enable users to receive real-time diagnosis of engine parts
through visual markers which also provide directions for maintenance.

6.6. Vehicle-to-vehicle communication

Vehicle-to-vehicle communication technology (V2V) operates as a revolutionary system

which brings substantial improvements for both safety levels and system reliability in vehicles.
Wireless V2V technology enables vehicles to share operating information such as speed and
position as well as acceleration status and road conditions live through their system networks
[214]. Through this wireless communication capability vehicles can identify future threats
beforehand so they collaboratively make decisions while minimising safety risks and
generating traffic flow efficiency gain. The V2V network alerts drivers about upcoming
vehicles that brake suddenly or warns about dangerous road conditions thus providing time for
drivers to take advance collision avoidance actions [215]. Scientists must concentrate on
building dependable and expansive V2V technologies which work effectively under diverse
real-world operating situations. V2V-related technologies must incorporate solutions that
address issues with fast data movement and network blocking, and resist environmental
changes including city tunnels and harsh climate situations [216]. Timely accurate transfer of
information from V2V systems must reach both drivers and all important stakeholders who
include fleet managers and traffic control systems. The integration of V2V communication
networks shows promise as a new advancement for VHM systems. V2V allows vehicles to
share data in real time thus enabling cooperative fault diagnosis through joint vehicle analysis
of potential errors. The joint effort between vehicles allows improved fault recognition together
with a complete understanding of vehicle health-status monitoring results. One vehicle that
notices an irregularity in its braking system shares this information through V2V
communication with surrounding vehicles so all nearby vehicles can modify their driving
decisions to prevent dangers [217].

6.7. Wireless monitoring

26
 The implementation of wireless monitoring systems in VHM proves vital because it enables
immediate data acquisition and dissemination through connectionless systems. Without the
requirement of invasive procedures, this technology makes it possible to smoothly connect
monitoring hardware including sensors and diagnostic tools to vehicles. Atlas systems illustrate
wireless monitoring hardware methods that monitor vehicle health indicators through
measurements that do not need physical component attachment [218]. VHM system
deployment becomes easier through wireless capabilities and these capabilities enhance
adaptations for multiple vehicle classes under varied operating conditions. Future scientists
developing wireless monitoring systems must design next-level wireless sensing systems able
to function with extreme operational conditions such as heat, motion and electromagnetic
disturbances to deliver dependable information transmission over large distances. In order to
quickly assess vehicle criticalities researchers must create wireless communication methods
and field-tested sensors that send information rapidly while maintaining low latency time [219].
For instance, advanced protocols like 5G and LoRaWAN can provide the necessary bandwidth
and range for transmitting large volumes of sensor data in real time, even in challenging
environments [220]. A key area of innovation is the integration of wireless monitoring with
edge AI technologies. By processing data locally at the edge (i.e., within the vehicle or nearby
infrastructure), edge AI can significantly reduce latency and improve the responsiveness of
VHM systems. This approach also helps conserve network bandwidth by transmitting only the
most critical data to centralised cloud systems for further analysis. For example, edge AI
algorithms can analyse sensor data in real time to detect anomalies and trigger immediate
responses, such as alerting the driver or initiating preventive maintenance actions [221].

7. Conclusions

Integrating advanced technologies and intelligent techniques has revolutionised vehicle

health monitoring and fault diagnosis in the automotive industry. This review consolidates
traditional, electric, and autonomous vehicle methodologies, demonstrating how sensors, IoT,
big data analytics, AI, and machine learning enable real-time monitoring, early fault detection,
and proactive maintenance.

1) Machine learning and deep learning algorithms have demonstrated superior

performance over traditional model-based and signal-based methods across all vehicle
types. AI-driven fault diagnosis in conventional vehicles achieves up to 95% accuracy
in the early detection of mechanical failures. In comparison, advanced battery health
monitoring techniques in electric vehicles improve energy efficiency by 15-20% and
extend battery lifespan by 10-15%. Sensor fusion for autonomous vehicles with AI-
based fault prediction, achieving 99% reliable real-time decision-making. The multi-
sensor data integration via IoT frameworks has resulted in quantifiable reductions in
maintenance costs and an unexpected decrease in vehicle downtimes. These techniques
make it possible to accurately and promptly identify failures even in complex dynamic
vehicle systems.
2) Integrating artificial intelligence with vehicular sensor networks has empowered
predictive systems capable of failure forecasting and remaining practical life
estimation for critical components. It even allows for preventive maintenance, helping
to cut unscheduled breakdowns by as much as 40% and streamlining maintenance
schedules. IoT-based multi-sensor data fusion is essential to predictive maintenance
systems and depends on data processing methods. Employing diverse data sources for
fusion has improved fault diagnosis accuracy, extended equipment lifetime and
resulted in up to 20-30% reduction in maintenance costs.

27
 3) Deep learning approaches based on multi-sensor information fusion have demonstrated
impressive results in fault classification and diagnosis tasks for complicated systems
like electric powertrains and autonomous driving systems. Proposed conceptual
framework for real-time health monitoring and fault diagnosis constructs a systematic
implementation methodology to tackle the risk of integrating data processing at scale
in the automotive industry.
4) The automotive industry benefits substantially from these technological innovations.
AI-driven systems strengthen critical engine units, transmission systems, and electric
powertrain components, reducing maintenance expenses and downtime. ML and IoT-
powered predictive maintenance systems enhance batteries’ and motors’ operational
life and performance, addressing degradation and efficiency challenges. These
technologies improve fleet management operations while providing logistics and
transportation companies with cost-effective solutions.
5) This review makes several novel contributions to the field, including a unified
taxonomic classification explicitly tailored to different vehicle types, an analysis of
technology integration patterns across the automotive spectrum, a practical conceptual
framework that accommodates heterogeneous vehicle platforms, and comparative
insights into the transferability of diagnostic techniques between conventional, electric,
and autonomous systems.
6) We recommend vehicle-specific implementation strategies: integrating AI diagnostic
systems into legacy vehicle architectures for conventional vehicles, developing secure
battery monitoring systems for EVs using fusion algorithms, and creating real-time
fault detection algorithms for AV sensor systems. Additional priorities include
standardising data protocols across diverse platforms, enhancing cybersecurity
frameworks for vehicular networks, developing intuitive human-machine interfaces,
optimising edge computing for real-time processing, fostering cross-disciplinary
collaboration, and advancing wireless monitoring systems for extreme operational
environments.
7) These research directions will enable transformative applications, including
cooperative fault diagnosis networks and lifecycle-optimised maintenance services.
The overarching impact of these advancements extends beyond operational efficiency
to encompass broader societal benefits: reduced environmental impact through
optimised vehicle performance, enhanced transportation safety, and increased
accessibility of reliable transportation options. By addressing immediate technical
challenges and sustainability goals, advanced VHM systems will play a crucial role in
shaping the future of safe, efficient, and sustainable transportation.

The frameworks presented in this review provide for future research and development in
vehicle health monitoring. By adopting these advanced techniques, the automotive industry
can achieve higher reliability, improved safety, and reduced operational costs while meeting
the evolving demands of sustainable and intelligent transportation systems.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.

28
Author contributions

Md Naeem Hossain: Conceptualisation, Methodology, Data curation, Formal analysis,

Writing- Original draft; Md Mustafizur Rahman: Conceptualisation, Writing- Review &
Editing, Project administration, Supervision; Devarajan Ramasamy: Investigation,
Visualisation, Funding acquisition.

Acknowledgements

The authors thank the Malaysia Ministry of Higher Education for supporting this work and
the University of Malaysia Pahang Al-Sultan Abdullah for its financial support. This work was
funded by the Fundamental Research Grant Scheme (No. FRGS/1/2022/TK10/UMP/02/35) of
the Ministry of Higher Education and the additional financial resources given through the
University Distinguished Research Grant (Project number RDU223016) to the Universiti
Malaysia Pahang Al-Sultan Abdullah.

References

[1] R. Danielis, M. Scorrano, M. Giansoldati, Decarbonising transport in Europe: Trends,

goals, policies and passenger car scenarios, Res. Transp. Econ. 91 (2022) 101068.
[2] E. Oh, M.M. Godoy Zúniga, T.B. Nguyen, B.-H. Kim, T. Trung Tien, J. Suhr,
Sustainable green composite materials in the next-generation mobility industry: review
and prospective, Adv. Compos. Mater. 33 (2024) 1368–1419.
[3] M. Oh, H. Moon, J. Woo, H. Kim, J. Shin, How to be a winner of future vehicle
maintenance services: consumer preference for vehicle self-diagnosis and fault
prediction system in next-generation vehicles, Technol. Anal. Strateg. Manag. (2024)
1–13.
[4] R.J. Shipley, B.A. Miller, R.J. Parrington, Introduction to failure analysis and prevention,
J. Fail. Anal. Prev. 22 (2022) 9–41.
[5] C. Wang, T. Ji, F. Mao, Z. Wang, Z. Li, Prognostics and health management system for
electric vehicles with a hierarchy fusion framework: Concepts, architectures, and
methods, Adv. Civ. Eng. 2021 (2021) 6685900.
[6] R. Raveendran, K.B. Devika, S.C. Subramanian, Brake fault identification and fault-
tolerant directional stability control of heavy road vehicles, IEEE Access 8 (2020)
169229–169246.
[7] J. Jiang, X. Cong, S. Li, C. Zhang, W. Zhang, Y. Jiang, A hybrid signal-based fault
diagnosis method for lithium-ion batteries in electric vehicles, IEEE Access 9 (2021)
19175–19186.
[8] S. Vasavi, K. Aswarth, T.S.D. Pavan, A.A. Gokhale, Predictive analytics as a service
for vehicle health monitoring using edge computing and AK-NN algorithm, Mater.
Today Proc. 46 (2021) 8645–8654.
[9] J. Lee, J. Ni, J. Singh, B. Jiang, M. Azamfar, J. Feng, Intelligent maintenance systems
and predictive manufacturing, J. Manuf. Sci. Eng. 142 (2020) 110805.
[10] M.N. Hossain, M.A. Rahim, M.M. Rahman, D. Ramasamy, Artificial intelligence
revolutionising the automotive sector: A comprehensive review of current insights,
challenges, and future scope, Comput. Mater. Contin. 82 (2025) 3643–3692.
[11] S. Barja-Martinez, M. Aragüés-Peñalba, Í. Munné-Collado, P. Lloret-Gallego, E.
Bullich-Massague, R. Villafafila-Robles, Artificial intelligence techniques for enabling
Big Data services in distribution networks: A review, Renew. Sustain. Energy Rev. 150
(2021) 111459.

29
[12] X. Xu, H. Li, W. Xu, Z. Liu, L. Yao, F. Dai, Artificial intelligence for edge service
optimization in internet of vehicles: A survey, Tsinghua Sci. Technol. 27 (2021) 270–
287.
[13] J.J. Eckert, F.M. Santiciolli, R.Y. Yamashita, F.C. Corrêa, L.C.A. Silva, F.G. Dedini,
Fuzzy gear shifting control optimisation to improve vehicle performance, fuel
consumption and engine emissions, IET Control Theory Appl. 13 (2019) 2658–2669.
[14] G. Niu, Data-driven technology for engineering systems health management, Springer
Singapore 10 (2017) 978–981.
[15] A. Choudhary, S. Fatima, B.K. Panigrahi, State of the art technologies in fault diagnosis
of electric vehicles: A component-based review, IEEE Trans. Transp. Electrif. 9 (2022)
2324–2347.
[16] X. Zhang, Y. Hu, J. Deng, H. Xu, H. Wen, Feature engineering and artificial
intelligence-supported approaches used for electric powertrain fault diagnosis: A review,
IEEE Access 10 (2022) 29069–29088.
[17] M.M. Abdeltwab, N.M. Ghazaly, A review on engine fault diagnosis through vibration
analysis, Int. J. Recent Technol. Mech. Electr. Eng. 9 (2022) 1–6.
[18] G. Anbazhagan, S. Jayakumar, S. Muthusamy, S.C.M. Sundararajan, H. Panchal, K.K.
Sadasivuni, An effective energy management strategy in hybrid electric vehicles using
Taguchi based approach for improved performance, Energy Sources, Part A Recover.
Util. Environ. Eff. 44 (2022) 3418–3435.
[19] G.N.R. Prasad, L.K. Kanulla, V. Ijjagiri, S.S.C. Mary, Implementation and health
monitoring system of vehicle by using IoT and cloud computing, 6th Int. Conf. Electron.
Commun. Aerosp. Technol. ICECA 2022 - Proc. (2022) 518–521.
[20] U. Shafi, A. Safi, A.R. Shahid, S. Ziauddin, M.Q. Saleem, Vehicle remote health
monitoring and prognostic maintenance system, J. Adv. Transp. 2018 (2018) 8061514.
[21] S. Vasavi, K. Aswarth, T. Sai Durga Pavan, A. Anu Gokhale, Predictive analytics as a
service for vehicle health monitoring using edge computing and AK-NN algorithm,
Mater. Today Proc. 46 (2021) 8645–8654.
[22] M.A. Rahim, M.A. Rahman, M.M. Rahman, N. Zaman, N. Moustafa, I. Razzak, An
intelligent risk management framework for monitoring vehicular engine health, IEEE
Trans. Green Commun. Netw. 6 (2022) 1298–1306.
[23] W. Xu, C. Yan, W. Jia, X. Ji, J. Liu, Analyzing and enhancing the security of ultrasonic
sensors for autonomous vehicles, IEEE Internet Things J. 5 (2018) 5015–5029.
[24] C. Du, X. Wei, Fault diagnosis of vehicle engine based on analytic hierarchy process
and neural network, in: 2018 Int. Conf. Mech. Electron. Control Autom. Eng. (MECAE
2018), Atlantis Press, 2018.
[25] A. Daas, B. Sari, J. Jia, G. Rigatos, Enhancing induction machine fault detection through
machine learning: Time and frequency analysis of vibration signals, Measurement 242
(2025) 116023.
[26] X. Zou, K. Zhang, T. Liu, Z. Jiang, Y. Xu, An overlapping group sparse variation
method for enhancing time–frequency modulation bispectrum characteristics and its
applications in bearing fault diagnosis, Measurement (2025) 117066.
[27] T. Praveenkumar, B. Sabhrish, M. Saimurugan, K.I. Ramachandran, Pattern recognition
based on-line vibration monitoring system for fault diagnosis of automobile gearbox,
Measurement 114 (2018) 233–242.
[28] G. Wang, S. Yin, Data-driven fault diagnosis for an automobile suspension system by
using a clustering based method, J. Franklin Inst. 351 (2014) 3231–3244.
[29] K. Wang, R. He, H. Li, J. Tang, R. Liu, Y. Li, P. He, Observer-based control for active
suspension system with time-varying delay and uncertainty, Adv. Mech. Eng. 11 (2019)
1687814019889505.

30
[30] R. Raveendran, K.B. Devika, S.C. Subramanian, Intelligent fault diagnosis of air brake
system in heavy commercial road vehicles, in: 2020 Int. Conf. Commun. Syst.
NETworkS, IEEE, 2020: pp. 93–98.
[31] X. Jin, G. Yin, N. Chen, Advanced estimation techniques for vehicle system dynamic
state: A survey, Sensors 19 (2019) 4289.
[32] B. Zhang, S. Lu, W. Xie, F. Xie, Fault detection and diagnosis based on interactive
multi-model moving horizon estimation and neuro-tire model, IEEE/ASME Trans.
Mechatronics 29 (2024) 3614–3625.
[33] V. Vasan, N.V. Sridharan, A. Prabhakaranpillai Sreelatha, S. Vaithiyanathan, Tire
condition monitoring using transfer learning-based deep neural network approach,
Sensors 23 (2023) 2177.
[34] M.Z. Khaneghah, M. Alzayed, H. Chaoui, Fault detection and diagnosis of the electric
motor drive and battery system of electric vehicles, Machines 11 (2023) 713.
[35] M. Schmid, H.-G. Kneidinger, C. Endisch, Data-driven fault diagnosis in battery
systems through cross-cell monitoring, IEEE Sens. J. 21 (2020) 1829–1837.
[36] L. Yao, Y. Xiao, X. Gong, J. Hou, X. Chen, A novel intelligent method for fault
diagnosis of electric vehicle battery system based on wavelet neural network, J. Power
Sources 453 (2020) 227870.
[37] H. Kaplan, K. Tehrani, M. Jamshidi, A fault diagnosis design based on deep learning
approach for electric vehicle applications, Energies 14 (2021) 6599.
[38] L. Gong, A. Sharma, M.A. Bhuiya, H. Awad, M.Z. Youssef, An adaptive fault diagnosis
of electric vehicles: An artificial intelligence blended signal processing methodology,
IEEE Can. J. Electr. Comput. Eng. 46 (2023) 196–206.
[39] Q. Shi, H. Zhang, Fault diagnosis of an autonomous vehicle with an improved SVM
algorithm subject to unbalanced datasets, IEEE Trans. Ind. Electron. 68 (2020) 6248–
6256.
[40] L. Biddle, S. Fallah, A novel fault detection, identification and prediction approach for
autonomous vehicle controllers using SVM, Automot. Innov. 4 (2021) 301–314.
[41] J. Ren, M. Green, X. Huang, From traditional to deep learning: Fault diagnosis for
autonomous vehicles, in: Learn. Control, Elsevier, 2021: pp. 205–219.
[42] S. Park, K. Oh, Y. Jeong, K. Yi, Model predictive control-based fault detection and
reconstruction algorithm for longitudinal control of autonomous driving vehicle using
multi-sliding mode observer, Microsyst. Technol. 26 (2020) 239–264.
[43] R. Bhagavathi, D.K.M. Kufoalor, A. Hasan, Digital twin-driven fault diagnosis for
autonomous surface vehicles, IEEE Access 11 (2023) 41096–41104.
[44] Y. Fang, C. Cheng, Z. Dong, H. Min, X. Zhao, A fault diagnosis framework for
autonomous vehicles based on hybrid data analysis methods combined with fuzzy PID
control, in: 2020 3rd Int. Conf. Unmanned Syst., IEEE, 2020: pp. 281–286.
[45] R. Yin, D. Wang, S. Zhao, Z. Lou, G. Shen, Wearable sensors‐enabled human–machine
interaction systems: from design to application, Adv. Funct. Mater. 31 (2021) 2008936.
[46] S. Urooj, F. Alrowais, Y. Teekaraman, H. Manoharan, R. Kuppusamy, IoT based
electric vehicle application using boosting algorithm for smart cities, Energies 14 (2021)
1072.
[47] R. Abachi, Overview of automotive sensors, Sensors (Peterborough, NH) 13 (1996) 82–
85.
[48] E.F. da Silva Neto, A.R.S. Feitosa, G.D.C. Cavalcanti, A.G. Silva-Filho, Detecting
anomalies in the engine coolant sensor using one-class classifiers, in: 2019 IEEE 90th
Veh. Technol. Conf., IEEE, 2019: pp. 1–5.
[49] P. Gierth, L. Rebenklau, K. Augsburg, E. Bachmann, L. Niedermeyer, Novel
thermocouples for automotive applications, J. Sensors Sens. Syst. 7 (2018) 43–49.

31
[50] C. Guardiola, B. Pla, P. Bares, A. Stefanopoulou, Cylinder charge composition
observation based on in-cylinder pressure measurement, Measurement 131 (2019) 559–
568.
[51] J. Qian, D.S. Kim, D.W. Lee, On-vehicle triboelectric nanogenerator enabled self-
powered sensor for tire pressure monitoring, Nano Energy 49 (2018) 126–136.
[52] M. Henningsson, P. Tunestål, R. Johansson, A virtual sensor for predicting diesel engine
emissions from cylinder pressure data, IFAC Proc. Vol. 45 (2012) 424–431.
[53] L. Chen, M. Mehregany, A silicon carbide capacitive pressure sensor for in-cylinder
pressure measurement, Sensors Actuators A Phys. 145–146 (2008) 2–8.
[54] M. Pavlin, F. Novak, Yield enhancement of piezoresistive pressure sensors for
automotive applications, Sensors Actuators A Phys. 141 (2008) 34–42.
[55] U.P. Borole, S. Subramaniam, I.R. Kulkarni, P. Saravanan, H.C. Barshilia, P.
Chowdhury, Highly sensitive giant magnetoresistance (GMR) based ultra low
differential pressure sensor, Sensors Actuators A Phys. 280 (2018) 125–131.
[56] S. Neupane, I.A. Fernandez, W. Patterson, S. Mittal, S. Rahimi, A temporal anomaly
detection system for vehicles utilizing functional working groups and sensor channels,
in: 2022 IEEE 8th Int. Conf. Collab. Internet Comput., 2022: pp. 99–108.
[57] T. Ritter, G. Hagen, J. Lattus, R. Moos, Solid state mixed-potential sensors as direct
conversion sensors for automotive catalysts, Sensors Actuators B Chem. 255 (2018)
3025–3032.
[58] U. Javed, K.P. Ramaiyan, C.R. Kreller, E.L. Brosha, R. Mukundan, A. V Morozov,
Using sensor arrays to decode NOx/NH3/C3H8 gas mixtures for automotive exhaust
monitoring, Sensors Actuators B Chem. 264 (2018) 110–118.
[59] S. Halley, K.P. Ramaiyan, L. Tsui, F. Garzon, A review of zirconia oxygen, NOx, and
mixed potential gas sensors-History and current trends, Sensors Actuators B Chem. 370
(2022) 132363.
[60] F. Gao, W. Xuan, A. Bermak, F. Boussaid, C.-Y. Tsui, J. Luo, Dual transduction on a
single sensor for gas identification, Sensors Actuators B Chem. 278 (2019) 21–27.
[61] K. Jo, K. Chu, M. Sunwoo, Interacting multiple model filter-based sensor fusion of GPS
with in-vehicle sensors for real-time vehicle positioning, IEEE Trans. Intell. Transp.
Syst. 13 (2011) 329–343.
[62] P. Mohankumar, J. Ajayan, R. Yasodharan, P. Devendran, R. Sambasivam, A review of
micromachined sensors for automotive applications, Measurement 140 (2019) 305–322.
[63] F. Pollicino, S. Eisa, P. Rosa, M.L. Pardal, M. Marchetti, Decentralized position
detection for moving vehicles, in: 2023 IEEE 97th Veh. Technol. Conf., IEEE, 2023:
pp. 1–6.
[64] D. Viramgama, Intelligent vehicle safety technology: Anti-lock braking system (ABS)
and its advancements, Int. Res. J. Eng. Technol. 6 (2019) 1158–1163.
[65] F. Abd Rahman, A.E. Gomes, N.A. Kamsani, R.M. Sidek, S.J. Hashim, F.Z. Rokhani,
M.S. Zamali, F.M. Ali, Random missing tooth error detection in crankshaft function of
an engine control unit, J. Mech. Eng. Sci. 14 (2020) 6895–6905.
[66] C. Chen, T. Ma, H. Jin, Y. Wu, Z. Hou, F. Li, Torque and rotational speed sensor based
on resistance and capacitive grating for rotational shaft of mechanical systems, Mech.
Syst. Signal Process. 142 (2020) 106737.
[67] L. Feng, W. Chen, T. Wu, H. Wang, D. Dai, D. Wang, W. Zhang, An improved sensor
system for wheel force detection with motion-force decoupling technique, Measurement
119 (2018) 205–217.
[68] M.S. Rauscher, M. Schardt, M.H. Köhler, A.W. Koch, Dual-channel mid-infrared
sensor based on tunable Fabry-Pérot filters for fluid monitoring applications, Sensors
Actuators B Chem. 259 (2018) 420–427.

32
[69] J. Mohammadpour, M. Franchek, K. Grigoriadis, A survey on diagnostic methods for
automotive engines, Int. J. Engine Res. 13 (2012) 41–64.
[70] T. Abukhalil, H. AlMahafzah, M. Alksasbeh, B.A.Y. Alqaralleh, Fuel consumption
using OBD-II and support vector machine model, J. Robot. 2020 (2020) 9450178.
[71] P. Tawadros, M. Awadallah, P. Walker, N. Zhang, Using a low-cost bluetooth torque
sensor for vehicle jerk and transient torque measurement, Proc. Inst. Mech. Eng. Part D
J. Automob. Eng. 234 (2020) 423–437.
[72] E. Pelletier, W. Bai, M.A. Tiburcio, J. Borek, S. Boyle, C. Earnhardt, L. Gao, S. Geyer,
C. Graham, B. Groelke, In-vehicle validation of heavy-duty vehicle fuel savings via a
hierarchical predictive online controller, SAE Int. J. Adv. Curr. Pract. Mobil. 3 (2021)
2159–2169.
[73] A. Mustayen, M.G. Rasul, X. Wang, M.A. Hazrat, M.I. Jahirul, M. Negnevitsky, Plastic-
made diesel (PMD) from pyrolysis via vacuum distillation process-A waste recycling
fuel to diesel engine performance and emissions improvement, J. Energy Inst. 107 (2023)
101198.
[74] K. Zhang, X. Hu, Y. Liu, X. Lin, W. Liu, Multi-fault detection and isolation for lithium-
ion battery systems, IEEE Trans. Power Electron. 37 (2021) 971–989.
[75] M. Lelie, T. Braun, M. Knips, H. Nordmann, F. Ringbeck, H. Zappen, D.U. Sauer,
Battery management system hardware concepts: An overview, Appl. Sci. 8 (2018) 534.
[76] F. Yang, W. Li, C. Li, Q. Miao, State-of-charge estimation of lithium-ion batteries based
on gated recurrent neural network, Energy 175 (2019) 66–75.
[77] A.K. Basu, S. Tatiya, S. Bhattacharya, Overview of electric vehicles (EVs) and EV
sensors, Sensors Automot. Aerosp. Appl. (2019) 107–122.
[78] P. Gangsar, R. Tiwari, Signal based condition monitoring techniques for fault detection
and diagnosis of induction motors: A state-of-the-art review, Mech. Syst. Signal Process.
144 (2020) 106908.
[79] B.N.C. V Chakravarthi, P. Naveen, S. Pragaspathy, V.S.N.N. Raju, Performance of
induction motor with hybrid multi level inverter for electric vehicles, in: 2021 Int. Conf.
Artif. Intell. Smart Syst., IEEE, 2021: pp. 1474–1478.
[80] Y. Li, H. Wu, X. Xu, X. Sun, J. Zhao, Rotor position estimation approaches for
sensorless control of permanent magnet traction motor in electric vehicles: A review,
World Electr. Veh. J. 12 (2021) 9.
[81] A. Brighente, M. Conti, D. Donadel, R. Poovendran, F. Turrin, J. Zhou, Electric vehicles
security and privacy: Challenges, solutions, and future needs, ArXiv Prepr.
ArXiv2301.04587 (2023).
[82] S. Hemavathi, A. Shinisha, A study on trends and developments in electric vehicle
charging technologies, J. Energy Storage 52 (2022) 105013.
[83] S.S. Kadlag, M.P. Thakre, R. Mapari, R. Shriwastava, P.C. Tapre, D.P. Kadam, A novel
pulse charger with intelligent battery management system for fast charging of electric
vehicle, Bull. Electr. Eng. Informatics 12 (2023) 1388–1396.
[84] X. Liu, C. Liu, W. Han, P.W.T. Pong, Design and implementation of a multi-purpose
TMR sensor matrix for wireless electric vehicle charging, IEEE Sens. J. 19 (2018) 1683–
1692.
[85] R. Palanisamy, R. Sahasrabuddhe, K.H. Mathur, J.A. Puranik, A new energy
regeneration system for a BLDC motor driven electric vehicle, Int. J. Electr. Comput.
Eng. 11 (2021) 2986.
[86] K.V. Singh, H.O. Bansal, D. Singh, A comprehensive review on hybrid electric vehicles:
architectures and components, J. Mod. Transp. 27 (2019) 77–107.
[87] Z. Zhang, Y. Dong, Y. Han, Dynamic and control of electric vehicle in regenerative
braking for driving safety and energy conservation, J. Vib. Eng. Technol. 8 (2020) 179–

33
 197.
[88] A.K. Sesha, L. Zhang, J. Mookken, IGBT power module with integrated current and
temperature sensor for HEV/EV traction inverter, in: PCIM Eur. 2019; Int. Exhib. Conf.
Power Electron. Intell. Motion, Renew. Energy Energy Manag., VDE, 2019: pp. 1–6.
[89] M. Anwar, M. Hayes, A. Tata, M. Teimorzadeh, T. Achatz, Power dense and robust
traction power inverter for the second-generation chevrolet volt extended-range EV,
SAE Int. J. Altern. Powertrains 4 (2015) 145–152.
[90] D. Boroyevich, I. Cvetković, D. Dong, R. Burgos, F. Wang, F. Lee, Future electronic
power distribution systems a contemplative view, in: 2010 12th Int. Conf. Optim. Electr.
Electron. Equip., IEEE, 2010: pp. 1369–1380.
[91] H. Liu, L. Han, Y. Cao, Improving transmission efficiency and reducing energy
consumption with automotive continuously variable transmission: A model prediction
comprehensive optimization approach, Appl. Energy 274 (2020) 115303.
[92] S.M.N. Sakib, LiDAR Technology-an overview, IUP J. Electr. Electron. Eng. 15 (2022)
36–57.
[93] J. Alfred Daniel, C. Chandru Vignesh, B.A. Muthu, R. Senthil Kumar, C.B.
Sivaparthipan, C.E.M. Marin, Fully convolutional neural networks for LIDAR–camera
fusion for pedestrian detection in autonomous vehicle, Multimed. Tools Appl. 82 (2023)
25107–25130.
[94] C. Cornwall, Event-Based Obstacle Detection with Commercial Lidar, (2022).
[95] J. Beltrán, C. Guindel, A. de la Escalera, F. García, Automatic extrinsic calibration
method for lidar and camera sensor setups, IEEE Trans. Intell. Transp. Syst. 23 (2022)
17677–17689.
[96] J. Schulte-Tigges, M. Förster, G. Nikolovski, M. Reke, A. Ferrein, D. Kaszner, D.
Matheis, T. Walter, Benchmarking of various LiDAR sensors for use in self-driving
vehicles in real-world environments, Sensors 22 (2022) 7146.
[97] R. Changalvala, H. Malik, LiDAR data integrity verification for autonomous vehicle,
IEEE Access 7 (2019) 138018–138031.
[98] S. Campbell, N. O’Mahony, L. Krpalcova, D. Riordan, J. Walsh, A. Murphy, C. Ryan,
Sensor technology in autonomous vehicles: A review, in: 2018 29th Irish Signals Syst.
Conf., IEEE, 2018: pp. 1–4.
[99] V.K. Kukkala, J. Tunnell, S. Pasricha, T. Bradley, Advanced driver-assistance systems:
A path toward autonomous vehicles, IEEE Consum. Electron. Mag. 7 (2018) 18–25.
[100] S. Wei, Y. Zou, X. Zhang, T. Zhang, X. Li, An integrated longitudinal and lateral vehicle
following control system with radar and vehicle-to-vehicle communication, IEEE Trans.
Veh. Technol. 68 (2019) 1116–1127.
[101] A. Stateczny, W. Kazimierski, D. Gronska-Sledz, W. Motyl, The empirical application
of automotive 3D radar sensor for target detection for an autonomous surface vehicle’s
navigation, Remote Sens. 11 (2019) 1156.
[102] H.A. Ignatious, M. Khan, An overview of sensors in autonomous vehicles, Procedia
Comput. Sci. 198 (2022) 736–741.
[103] W. Shi, M.B. Alawieh, X. Li, H. Yu, Algorithm and hardware implementation for visual
perception system in autonomous vehicle: A survey, Integration 59 (2017) 148–156.
[104] J. Anitha, G. Mani, K. Venkata Rao, Driver drowsiness detection using viola jones
algorithm, in: Smart Intell. Comput. Appl. Proc. Third Int. Conf. Smart Comput.
Informatics, Vol. 1, Springer, 2020: pp. 583–592.
[105] A. Barodi, A. Bajit, M. Benbrahim, A. Tamtaoui, An enhanced approach in detecting
object applied to automotive traffic roads signs, in: 2020 IEEE 6th Int. Conf. Optim.
Appl., IEEE, 2020: pp. 1–6.
[106] F. Chang, M. Liu, M. Dong, Y. Duan, A mobile vision inspection system for tiny defect

34
 detection on smooth car-body surfaces based on deep ensemble learning, Meas. Sci.
Technol. 30 (2019) 125905.
[107] M. Toa, A. Whitehead, Ultrasonic sensing Basics, Dallas: Texas Instruments (2020).
[108] L. Zhu, Analyze the advantages and disadvantages of different sensors for autonomous
vehicles, in: 2022 7th Int. Conf. Soc. Sci. Econ. Dev. (ICSSED 2022), Atlantis Press,
2022: pp. 1020–1024.
[109] A. Aliyu, J.G. Kolo, O.O. Mikail, J. Agajo, B. Umar, O.I. Aguagba, An ultrasonic sensor
distance induced automatic braking automobile collision avoidance system, in: 2017
IEEE 3rd Int. Conf. Electro-Technology Natl. Dev., IEEE, 2017: pp. 570–576.
[110] B. Khaleghi, A. Khamis, F.O. Karray, S.N. Razavi, Multisensor data fusion: A review
of the state-of-the-art, Inf. Fusion 14 (2013) 28–44.
[111] D.J. Yeong, G. Velasco-Hernandez, J. Barry, J. Walsh, Sensor and sensor fusion
technology in autonomous vehicles: A review, Sensors 21 (2021) 2140.
[112] S. Qiu, H. Zhao, N. Jiang, Z. Wang, L. Liu, Y. An, H. Zhao, X. Miao, R. Liu, G. Fortino,
Multi-sensor information fusion based on machine learning for real applications in
human activity recognition: State-of-the-art and research challenges, Inf. Fusion 80
(2022) 241–265.
[113] T. Meng, X. Jing, Z. Yan, W. Pedrycz, A survey on machine learning for data fusion,
Inf. Fusion 57 (2020) 115–129.
[114] Ö. Gültekin, E. Cinar, K. Özkan, A. Yazıcı, Multisensory data fusion-based deep
learning approach for fault diagnosis of an industrial autonomous transfer vehicle,
Expert Syst. Appl. 200 (2022) 117055.
[115] T.P. Kumar, M. Saimurugan, R.B.H. Haran, S. Siddharth, K.I. Ramachandran, A multi-
sensor information fusion for fault diagnosis of a gearbox utilizing discrete wavelet
features, Meas. Sci. Technol. 30 (2019) 85101.
[116] M. Yue, Z. Al Masry, S. Jemei, N. Zerhouni, An online prognostics-based health
management strategy for fuel cell hybrid electric vehicles, Int. J. Hydrogen Energy 46
(2021) 13206–13218.
[117] S. Wang, Y. Hou, F. Gao, X. Ji, A novel IoT access architecture for vehicle monitoring
system, in: 2016 IEEE 3rd World Forum Internet Things, IEEE, 2016: pp. 639–642.
[118] S. Phuyal, D. Bista, R. Bista, Challenges, opportunities and future directions of smart
manufacturing: a state of art review, Sustain. Futur. 2 (2020) 100023.
[119] H. Mrabet, S. Belguith, A. Alhomoud, A. Jemai, A survey of IoT security based on a
layered architecture of sensing and data analysis, Sensors 20 (2020) 3625.
[120] R. Krishnamurthi, A. Kumar, D. Gopinathan, A. Nayyar, B. Qureshi, An overview of
IoT sensor data processing, fusion, and analysis techniques, Sensors 20 (2020) 6076.
[121] V.M. Tayur, R. Suchithra, Review of interoperability approaches in application layer of
Internet of Things, in: 2017 Int. Conf. Innov. Mech. Ind. Appl., IEEE, 2017: pp. 322–
326.
[122] K.M. Alam, M. Saini, A. El Saddik, Toward social internet of vehicles: Concept,
architecture, and applications, IEEE Access 3 (2015) 343–357.
[123] B. Rezaeianjouybari, Y. Shang, Deep learning for prognostics and health management:
State of the art, challenges, and opportunities, Measurement 163 (2020) 107929.
[124] J. Wang, Y. Mo, Online monitoring of automotive engine lubricating oil based on
Internet of Things technology, Wirel. Commun. Mob. Comput. 2022 (2022) 2478186.
[125] Y. Jeong, S. Son, E. Jeong, B. Lee, An integrated self-diagnosis system for an
autonomous vehicle based on an IoT gateway and deep learning, Appl. Sci. 8 (2018)
1164.
[126] Y. Chi, Y. Dong, Z.J. Wang, F.R. Yu, V.C.M. Leung, Knowledge-based fault diagnosis
in industrial Internet of Things: A survey, IEEE Internet Things J. 9 (2022) 12886–

35
 12900.
[127] Y. Wen, Z. Zhu, Y. Dai, J. Hu, C. Chen, The research of intelligent fault diagnosis
system based on Internet of Things, in: 2019 Int. Conf. Internet Things IEEE Green
Comput. Commun. IEEE Cyber, Phys. Soc. Comput. IEEE Smart Data, IEEE, 2019: pp.
1120–1124.
[128] C. Dashora, P.E. Sudhagar, J. Marietta, IoT based framework for the detection of vehicle
accident, Cluster Comput. 23 (2020) 1235–1250.
[129] O. Benedettini, T.S. Baines, H.W. Lightfoot, R.M. Greenough, State-of-the-art in
integrated vehicle health management, Proc. Inst. Mech. Eng. Part G J. Aerosp. Eng.
223 (2009) 157–170.
[130] F. Arena, M. Collotta, L. Luca, M. Ruggieri, F.G. Termine, Predictive maintenance in
the automotive sector: A literature review, Math. Comput. Appl. 27 (2021) 2.
[131] A. Theissler, J. Pérez-Velázquez, M. Kettelgerdes, G. Elger, Predictive maintenance
enabled by machine learning: Use cases and challenges in the automotive industry,
Reliab. Eng. Syst. Saf. 215 (2021) 107864.
[132] T.K. Hua, A Short Review on Machine Learning, Authorea Prepr. (2022).
[133] W. Tong, A. Hussain, W.X. Bo, S. Maharjan, Artificial intelligence for vehicle-to-
everything: A survey, IEEE Access 7 (2019) 10823–10843.
[134] C.-S.A. Gong, C.-H.S. Su, Y.-H. Chen, D.-Y. Guu, How to implement automotive fault
diagnosis using artificial intelligence scheme, Micromachines 13 (2022) 1380.
[135] C.S.A. Gong, C.H.S. Su, K.H. Tseng, Implementation of machine learning for fault
classification on vehicle power transmission system, IEEE Sens. J. 20 (2020) 15163–
15176.
[136] X. Yan, M. Sarkar, B. Lartey, B. Gebru, A. Homaifar, A. Karimoddini, E. Tunstel, An
online learning framework for sensor fault diagnosis analysis in autonomous cars, IEEE
Trans. Intell. Transp. Syst. 24 (2023) 14467–14479.
[137] T. Mahmood, J. Li, Y. Pei, F. Akhtar, S.A. Butt, A. Ditta, S. Qureshi, An intelligent fault
detection approach based on reinforcement learning system in wireless sensor network,
J. Supercomput. 78 (2022) 3646–3675.
[138] N. Ahmed, R. Amin, H. Aldabbas, D. Koundal, B. Alouffi, T. Shah, Machine learning
techniques for spam detection in email and IoT platforms: Analysis and research
challenges, Secur. Commun. Networks 2022 (2022) 1862888.
[139] K.K. Hiran, R.K. Jain, K. Lakhwani, R. Doshi, Machine learning: Master supervised
and unsupervised learning algorithms with real examples (English Edition), BPB
Publications, 2021.
[140] E. Akanksha, N. Sharma, K. Gulati, Review on reinforcement learning, research
evolution and scope of application, in: 2021 5th Int. Conf. Comput. Methodol. Commun.,
IEEE, 2021: pp. 1416–1423.
[141] M. Elsisi, K. Mahmoud, M. Lehtonen, M.M.F. Darwish, Reliable industry 4.0 based on
machine learning and IOT for analyzing, monitoring, and securing smart meters, Sensors
21 (2021) 487.
[142] P. Ongsulee, Artificial intelligence, machine learning and deep learning, in: 2017 15th
Int. Conf. ICT Knowl. Eng. (ICT&KE), 2017; Bangkok, Thail., IEEE, 2017: pp. 1–6.
[143] I.H. Sarker, Machine learning: Algorithms, real-world applications and research
directions, SN Comput. Sci. 2 (2021) 160.
[144] L. Kaiser, M. Babaeizadeh, P. Milos, B. Osinski, R.H. Campbell, K. Czechowski, D.
Erhan, C. Finn, P. Kozakowski, S. Levine, Model-based reinforcement learning for atari,
ArXiv Prepr. ArXiv1903.00374 (2019).
[145] J.S.L. Senanayaka, H. Van Khang, K.G. Robbersmyr, Toward self-supervised feature
learning for online diagnosis of multiple faults in electric powertrains, IEEE Trans. Ind.

36
 Informatics 17 (2020) 3772–3781.
[146] C.-S.A. Gong, C.-H.S. Su, Y.-E. Liu, D.-Y. Guu, Y.-H. Chen, Deep learning with LPC
and wavelet algorithms for driving fault diagnosis, Sensors 22 (2022) 7072.
[147] Y. Echevarría, L. Sánchez, C. Blanco, Hybrid artificial intelligent systems, Lect. Notes
Comput. Sci. (Including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics)
9648 (2016) 250–260.
[148] G. San Martin, E. López Droguett, V. Meruane, M. das Chagas Moura, Deep variational
auto-encoders: A promising tool for dimensionality reduction and ball bearing elements
fault diagnosis, Struct. Heal. Monit. 18 (2019) 1092–1128.
[149] G. Li, J. Wu, C. Deng, X. Xu, X. Shao, Deep reinforcement learning-based online
domain adaptation method for fault diagnosis of rotating machinery, IEEE/ASME Trans.
Mechatronics 27 (2021) 2796–2805.
[150] M.O. Mahfooth, O.M. Mahmood, Detection and diagnosis of inter-turn short circuit
faults of PMSM for electric vehicles based on deep reinforcement learning, Al-Rafidain
Eng. J. 28 (2023) 75–85.
[151] H. Chen, Z. Liu, C. Alippi, B. Huang, D. Liu, Explainable intelligent fault diagnosis for
nonlinear dynamic systems: From unsupervised to supervised learning, IEEE Trans.
Neural Networks Learn. Syst. 35 (2022) 6166–6179.
[152] X. Zhao, M. Jia, A novel unsupervised deep learning network for intelligent fault
diagnosis of rotating machinery, Struct. Heal. Monit. 19 (2020) 1745–1763.
[153] A. Folkers, M. Rick, C. Büskens, Controlling an autonomous vehicle with deep
reinforcement learning, in: 2019 IEEE Intell. Veh. Symp., IEEE, 2019: pp. 2025–2031.
[154] M. Morales, Grokking deep reinforcement learning, Manning Publications, 2020.
[155] A. Abdelfattah, H. Ibrahim, Health monitoring of automotive suspension system using
machine learning, in: Int. Conf. Veh. Technol. Intell. Transp. Syst. VEHITS - Proc.,
2021: pp. 325–332.
[156] S. Ayesha, M.K. Hanif, R. Talib, Overview and comparative study of dimensionality
reduction techniques for high dimensional data, Inf. Fusion 59 (2020) 44–58.
[157] Y. Ding, L. Ma, J. Ma, M. Suo, L. Tao, Y. Cheng, C. Lu, Intelligent fault diagnosis for
rotating machinery using deep Q-network based health state classification: A deep
reinforcement learning approach, Adv. Eng. Informatics 42 (2019) 100977.
[158] A.M. Zador, A critique of pure learning and what artificial neural networks can learn
from animal brains, Nat. Commun. 10 (2019) 3770.
[159] H. Kukreja, N. Bharath, C.S. Siddesh, S. Kuldeep, An introduction to artificial neural
network, Int J Adv Res Innov Ideas Educ 1 (2016) 27–30.
[160] K. Su, J. Zhang, J. Zhang, T. Yan, G. Mei, Optimisation of current collection quality of
high-speed pantograph-catenary system using the combination of artificial neural
network and genetic algorithm, Veh. Syst. Dyn. 61 (2023) 260–285.
[161] H. Xue, M. Wu, Z. Zhang, H. Wang, Intelligent diagnosis of mechanical faults of in-
wheel motor based on improved artificial hydrocarbon networks, ISA Trans. 120 (2022)
360–371.
[162] C. Du, W. Li, Y. Rong, F. Li, F. Yu, X. Zeng, Research on the application of artificial
intelligence method in automobile engine fault diagnosis, Eng. Res. Express 3 (2021)
26002.
[163] L. Deng, Y. Liu, A joint introduction to natural language processing and to deep learning,
Deep Learn. Nat. Lang. Process. (2018) 1–22.
[164] J. Liu, A. Khattak, Informed decision-making by integrating historical on-road driving
performance data in high-resolution maps for connected and automated vehicles, J. Intell.
Transp. Syst. 24 (2020) 11–23.
[165] M. Al-Zeyadi, J. Andreu-Perez, H. Hagras, C. Royce, D. Smith, P. Rzonsowski, A.

37
 Malik, Deep learning towards intelligent vehicle fault diagnosis, in: Proc. Int. Jt. Conf.
Neural Networks, Institute of Electrical and Electronics Engineers Inc., 2020.
[166] C. Chen, P. Zhang, H. Zhang, J. Dai, Y. Yi, H. Zhang, Y. Zhang, Deep learning on
computational-resource-limited platforms: a survey, Mob. Inf. Syst. 2020 (2020) 1–19.
[167] T. Han, C. Liu, W. Yang, D. Jiang, A novel adversarial learning framework in deep
convolutional neural network for intelligent diagnosis of mechanical faults, Knowledge-
Based Syst. 165 (2019) 474–487.
[168] S. Ma, F. Chu, Ensemble deep learning-based fault diagnosis of rotor bearing systems,
Comput. Ind. 105 (2019) 143–152.
[169] H. Luo, M. Huang, Z. Zhou, A dual-tree complex wavelet enhanced convolutional
LSTM neural network for structural health monitoring of automotive suspension,
Measurement 137 (2019) 14–27.
[170] M. Alakhras, M. Oussalah, M. Hussein, A survey of fuzzy logic in wireless localization,
EURASIP J. Wirel. Commun. Netw. 2020 (2020) 89.
[171] J. Pan, A.T. Nguyen, T.M. Guerra, C. Sentouh, S. Wang, J.C. Popieul, Vehicle actuator
fault detection with finite-frequency specifications via Takagi-Sugeno fuzzy observers:
Theory and experiments, IEEE Trans. Veh. Technol. 72 (2022) 407–417.
[172] N. Mayadevi, V.P. Mini, R. Hari Kumar, S. Prins, Fuzzy-based intelligent algorithm for
diagnosis of drive faults in induction motor drive system, Arab. J. Sci. Eng. 45 (2020)
1385–1395.
[173] H.T. Nguyen, C.L. Walker, E.A. Walker, A first course in fuzzy logic, Chapman and
Hall/CRC., 2018.
[174] C.M. Ezhilarasu, Z. Skaf, I.K. Jennions, The application of reasoning to aerospace
Integrated Vehicle Health Management (IVHM): Challenges and opportunities, Prog.
Aerosp. Sci. 105 (2019) 60–73.
[175] X. Xu, D. Cao, Y. Zhou, J. Gao, Application of neural network algorithm in fault
diagnosis of mechanical intelligence, Mech. Syst. Signal Process. 141 (2020) 106625.
[176] M. Jiang, J. Lu, Maritime accident risk estimation for sea lanes based on a dynamic
Bayesian network, Marit. Policy Manag. 47 (2020) 649–664.
[177] Y. Li, Y. Liu, T. Huang, H. Huang, J. Mi, Reliability assessment for systems suffering
common cause failure based on Bayesian networks and proportional hazards model,
Qual. Reliab. Eng. Int. 36 (2020) 2509–2520.
[178] I.P. Gomes, D.F. Wolf, Health monitoring system for autonomous vehicles using
dynamic Bayesian networks for diagnosis and prognosis, J. Intell. Robot. Syst. 101
(2021) 19.
[179] H. Xue, J. Zhou, Z. Chen, Z. Li, Real-time diagnosis of an in-wheel motor of an electric
vehicle based on dynamic Bayesian networks, IEEE Access 7 (2019) 114685–114699.
[180] J. Wang, Z. Wang, V. Stetsyuk, X. Ma, F. Gu, W. Li, Exploiting Bayesian networks for
fault isolation: A diagnostic case study of diesel fuel injection system, ISA Trans. 86
(2019) 276–286.
[181] T. Bodrumlu, M.M. Gözüm, B. Kavak, Enhanced fault detection of vehicle lateral
dynamics using a dynamically adjustable Bayesian network structure and extented
Kalman filter, in: ASME Int. Mech. Eng. Congr. Expo., American Society of
Mechanical Engineers, 2022: p. V009T14A024.
[182] M.M.H. Shuvo, S.K. Islam, J. Cheng, B.I. Morshed, Efficient acceleration of deep
learning inference on resource-constrained edge devices: A review, Proc. IEEE 111
(2022) 42–91.
[183] H. Li, T. Wang, G. Wu, A Bayesian deep learning approach for random vibration
analysis of bridges subjected to vehicle dynamic interaction, Mech. Syst. Signal Process.
170 (2022) 108799.

38
[184] D. Oladimeji, K. Gupta, N.A. Kose, K. Gundogan, L. Ge, F. Liang, Smart transportation:
An overview of technologies and applications, Sensors 23 (2023) 3880.
[185] J. Wu, Z. Huang, C. Lv, Uncertainty-aware model-based reinforcement learning:
Methodology and application in autonomous driving, IEEE Trans. Intell. Veh. 8 (2022)
194–203.
[186] Y. Lu, H. Ma, E. Smart, H. Yu, Real-time performance-focused localization techniques
for autonomous vehicle: A review, IEEE Trans. Intell. Transp. Syst. 23 (2021) 6082–
6100.
[187] S. Safavi, M.A. Safavi, H. Hamid, S. Fallah, Multi-sensor fault detection, identification,
isolation and health forecasting for autonomous vehicles, Sensors 21 (2021) 2547.
[188] V.S.R. Kosuru, A. Kavasseri Venkitaraman, A smart battery management system for
electric vehicles using deep learning-based sensor fault detection, World Electr. Veh. J.
14 (2023) 101.
[189] F.P. García Márquez, P.J. Bernalte Sánchez, I. Segovia Ramírez, Acoustic inspection
system with unmanned aerial vehicles for wind turbines structure health monitoring,
Struct. Heal. Monit. 21 (2022) 485–500.
[190] C.S.A. Gong, C.H.S. Su, Y.H. Chen, D.Y. Guu, How to implement automotive fault
diagnosis using artificial intelligence scheme, Micromachines 13 (2022) 1380.
[191] Y. Ding, L. Ma, J. Ma, M. Suo, L. Tao, Y. Cheng, C. Lu, Intelligent fault diagnosis for
rotating machinery using deep Q-network based health state classification: A deep
reinforcement learning approach, Adv. Eng. Informatics 42 (2019) 100977.
[192] T. Nowakowski, P. Komorski, Diagnostics of the drive shaft bearing based on vibrations
in the high-frequency range as a part of the vehicle’s self-diagnostic system, Eksploat. i
Niezawodn. 24 (2022) 70–79.
[193] P. Jieyang, A. Kimmig, W. Dongkun, Z. Niu, F. Zhi, W. Jiahai, X. Liu, J. Ovtcharova,
A systematic review of data-driven approaches to fault diagnosis and early warning, J.
Intell. Manuf. 34 (2023) 3277–3304.
[194] G.W. Vogl, B.A. Weiss, M. Helu, A review of diagnostic and prognostic capabilities
and best practices for manufacturing, J. Intell. Manuf. 30 (2019) 79–95.
[195] K. Ranasinghe, R. Kapoor, A. Gardi, R. Sabatini, V. Wickramanayake, D. Ludovici,
Vehicular sensor network and data analytics for a health and usage management system,
Sensors 20 (2020) 5892.
[196] G. Reiter, K. Williams, AI-based signal integrity monitoring for integrated vehicle
health management, in: ELIV 2021, VDI Wissensforum GmbH (ed.), Germany, 2021:
pp. 143–150.
[197] Ö. Aslan, S.S. Aktuğ, M. Ozkan-Okay, A.A. Yilmaz, E. Akin, A comprehensive review
of cyber security vulnerabilities, threats, attacks, and solutions, Electronics 12 (2023)
1333.
[198] N.M.A. Chisty, P.R. Baddam, R. Amin, Strategic approaches to safeguarding the digital
future: insights into next-generation cybersecurity, Eng. Int. 10 (2022) 69–84.
[199] S. Holland, Unsettled technology opportunities for vehicle health management and the
role for health-ready components, USA: SAE International; 2020. 32 p. SAE Technical
Paper EPR2020003, 2020.
[200] M. Malik, M. Maleki, P. Folkesson, B. Sangchoolie, J. Karlsson, ComFASE: A tool for
evaluating the effects of V2V communication faults and attacks on automated vehicles,
in: 2022 52nd Annu. IEEE/IFIP Int. Conf. Dependable Syst. Networks, IEEE, 2022: pp.
185–192.
[201] M. Yang, J. Zhang, Data anomaly detection in the internet of things: A review of current
trends and research challenges, Int. J. Adv. Comput. Sci. Appl. 14 (2023).
[202] M. Ahwiadi, W. Wang, Battery health monitoring and remaining useful life prediction

39
 techniques: A review of technologies, Batteries 11 (2025) 31.
[203] M.N. Hossain, M.M. Rahman, D. Ramasamy, Artificial intelligence-driven vehicle fault
diagnosis to revolutionize automotive maintenance: A review., C. Model. Eng. Sci. 141
(2024) 951–996.
[204] T. Sutharssan, S. Stoyanov, C. Bailey, C. Yin, Prognostic and health management for
engineering systems: a review of the data‐driven approach and algorithms, J. Eng. 2015
(2015) 215–222.
[205] J. Ren, C. Xu, J. Wang, J. Zhang, X. Mao, W. Shen, An edge-fog-cloud computing-
based digital twin model for prognostics health management of process manufacturing
systems., C. Model. Eng. Sci. 135 (2023) 599–618.
[206] M. Sadaf, Z. Iqbal, A.R. Javed, I. Saba, M. Krichen, S. Majeed, A. Raza, Connected and
automated vehicles: Infrastructure, applications, security, critical challenges, and future
aspects, Technologies 11 (2023) 117.
[207] P. Koopman, M. Wagner, Challenges in autonomous vehicle testing and validation, SAE
Int. J. Transp. Saf. 4 (2016) 15–24.
[208] K. Kuru, Trustfsdv: Framework for building and maintaining trust in self-driving
vehicles, IEEE Access 10 (2022) 82814–82833.
[209] J. Xing, D. Wei, S. Zhou, T. Wang, Y. Huang, H. Chen, A comprehensive study on self-
learning methods and implications to autonomous driving, IEEE Trans. Neural
Networks Learn. Syst. (2024) 1–20.
[210] M. Wilbrink, M. Lau, J. Illgner, A. Schieben, M. Oehl, Impact of external human–
machine interface communication strategies of automated vehicles on pedestrians’
crossing decisions and behaviors in an urban environment, Sustainability 13 (2021)
8396.
[211] S. Debernard, C. Chauvin, R. Pokam, S. Langlois, Designing human-machine interface
for autonomous vehicles, IFAC-PapersOnLine 49 (2016) 609–614.
[212] F. Tufano, S.W. Bahadure, M. Tufo, L. Novella, G. Fiengo, S. Santini, An optimization
framework for information management in adaptive automotive human–machine
interfaces, Appl. Sci. 13 (2023) 10687.
[213] Z. Zhang, F. Wen, Z. Sun, X. Guo, T. He, C. Lee, Artificial intelligence‐enabled sensing
technologies in the 5G/internet of things era: from virtual reality/augmented reality to
the digital twin, Adv. Intell. Syst. 4 (2022) 2100228.
[214] M.N. Hossain, M.M. Rahman, D. Ramasamy, K. Kadirgama, M.M. Noor,
Advancements, challenges, and implications for navigating the autonomous vehicle
revolution, J. Mech. Eng. Sci. 8 (2024) 10077–10093.
[215] S. Zeadally, J. Guerrero, J. Contreras, A tutorial survey on vehicle-to-vehicle
communications, Telecommun. Syst. 73 (2020) 469–489.
[216] M.N. Hossain, M.M. Rahman, D. Ramasamy, K. Kadirgama, M. Mat Noor, Progress ,
risks and impacts of the evolution of autonomous vehicles, J. Adv. Res. Appl. Sci. Eng.
Technol. 63 (2025) 97–118.
[217] D. Shin, B. Kim, K. Yi, A. Carvalho, F. Borrelli, Human-centered risk assessment of an
automated vehicle using vehicular wireless communication, IEEE Trans. Intell. Transp.
Syst. 20 (2018) 667–681.
[218] U. Ahmad, M. Han, A. Jolfaei, S. Jabbar, M. Ibrar, A. Erbad, H.H. Song, Y. Alkhrijah,
A comprehensive survey and tutorial on smart vehicles: Emerging technologies, security
issues, and solutions using machine learning, IEEE Trans. Intell. Transp. Syst. 25 (2024)
15314–15341.
[219] J. Li, Z. Wang, J. Hu, H. Liu, X. Yi, Communication-protocol-based distributed filtering
for general systems over sensor networks: developments and challenges, Int. J. Gen.
Syst. 51 (2022) 608–630.

40
[220] M.A.M. Almuhaya, W.A. Jabbar, N. Sulaiman, S. Abdulmalek, A survey on Lorawan
technology: Recent trends, opportunities, simulation tools and future directions,
Electronics 11 (2022) 164.
[221] Ö. Gültekin, E. Cinar, K. Özkan, A. Yazıcı, Real-time fault detection and condition
monitoring for industrial autonomous transfer vehicles utilizing edge artificial
intelligence, Sensors 22 (2022) 3208.

List of acronyms
Acronym Full form
AI Artificial intelligence
ANN Artificial neural network
AR Augmented reality
AV Autonomous vehicle
CNN Convolutional neural network
CV Conventional vehicle
DBN Dynamic Bayesian network
DL Deep learning
DNN Deep neural network
DT Decision tree
EKF Extended Kalman filter
EV Electric vehicle
HMI Human-machine interface
IoT Internet of Things
KNN K-nearest neighbour
LiDAR Light detection and ranging
LSTM Long short-term memory
ML Machine learning
MLP Multilayer perceptron
PCA Principal component analysis
PHM Prognostics and health management
PID Proportional-integral-derivative
RADAR Radio detection and ranging
RF Random forest
RMS Root mean square
RUL Remaining useful life
SOC State of charge
SVM Support vector machine
VHM Vehicle health monitoring
V2V Vehicle-to-vehicle

41
 Fig. 1. VHM correlations with sensors, AI, IoT, and others

Notes:
Cluster 1: VHM systems correlation Cluster 2: Fault diagnosis correlation
Cluster 3: Artificial intelligence networks Cluster 4: Sensors and sensor fusion
Cluster 5: Prognostics & accident prevention Cluster 6: Neural network correlation

42
Fig. 2. Taxonomy of VHM and fault diagnosis review

43
 Engine
Monitoring Techniques
ANN, Big Data, DT, Fuzzy
Monitoring Techniques
Logic, KNN, MLP, Naive
Artificial Immune Bayes, OBD, Practical Swarm
Recognition System, DT, Ensemble Optimization - Probabilistic
Learning, Hoeffding Trees, Multi- Neural Network, RF, SVM
Sensor Fusion, RF, Rough Set Battery Voltage, Coolant
Theory, SVM, Wavelets Temperature, Emission Level of
Exhaust Gases, Fuel Level, Load Monitoring Techniques
Variation, Mileage ANN, CNN, Generative
Adversarial Network, KNN,
Naive Bayes, RF, SVM

Brake Tire

Monitoring Parameters
Temperature Increase Rate, for Different Vehicular Inflation Pressure, Physical
Wheel Speed Sensor Data Systems Properties, Yaw Angle

Signal-based Stochastic Subspace

Identification, Linear Parameter
Gearbox Suspension

ANN, Average Correlation

Varying, LSTM, RF, SVM

Monitoring Techniques
Forest, Extreme Learning
Monitoring Techniques

Machine, MLP, SVM

CNN, Deep Random

Acoustic Emission, Controller Gains, Damping,

Accelerometer Measurement, Stiffness, State Residuals,
Gear Mesh Stiffness, Tire Radial Stiffness,
Rotational Speed Vehicle Body Mass

Fig. 3. Applied techniques for monitoring various parameters of a conventional vehicle

44
 Electric Vehicle Health
Monitoring and Fault
Diagnosis

Monitoring Parameters

Battery Management Electric Motor and

System Drive System

Voltage Temperature Time and Frequency Domain Symptom

State of Health State of Charge Voltage and Current Analysis

Sensor and Actuator Faults Temperature Measurement

Applied Techniques
Thermal Management System Faults Statistical Analysis

Model-based Systems Motor Current Signature Analysis

Model-based Systems Integrated Approach
Acoustic Signature Analysis Vibration-based Methods
Real-Time Voltage-based Method Signal Processing & Statistical Methods
ML & DL Techniques:
ML & DL Techniques: SVM, KNN, DT, PCA, CNN, DNN, AHN
Wavelet Neural Network, ELM, RF, LSTM, PCA
Hybrid and Ensemble Method

Fig. 4. Applied techniques for monitoring various parameters of electric vehicle

45
 RADAR Ultrasonic LiDAR Camera Charging Sensor Inverter Sensor

Speed Sensor Position Sensor

Current Temperatute

Autonomous Vehicle

Humidity Voltage
Temp. Sensor Gas Sensor Battery Sensors

Rotor Position
Torque Sensor

Stator Temp.
Pressure Sensor Motor Sensors
Electric Vehicle
Conventional Vehicle

Fig. 5. Illustration of sensors used in CVs, EVs, and AVs

46
 Feature Extraction Sensor 1 Feature Extraction

Decision Analysis Feature Extraction Sensor 2 Feature Extraction Feature Fusion

Decision Fusion Feature Extraction Sensor N Feature Extraction Decision Analysis

Decision Fusion Feature Fusion

Data Fusion

Data Fusion
Feature Extraction

Decision Analysis

Fig. 6. Multi-sensor fusion framework for VHM and fault diagnosis

47
 Network Layer Data Management
WiFi Bluetooth 5G Zigbee System

Processing Layer
Analytical
Information about Platform
Information about current, Video information, position,
temperature, pressure,
temperature, position, etc. environment, etc.
vibration, etc.

Interaction with the stakeholders

ML Algorithm through user friendly interface
about vehicle fitness

Battery Sensor,
Engine Sensor Radar, Camera
Motor Sensor
Sensing Layer

Driver

Application Layer

Conventional Autonomous
Electric Vehicle
Vehicle Vehicle
Function Apps

Fig. 7. IoT architecture for VHM and fault diagnosis system

48
 Communication Protocols

Data Filtering
Individual Sensor Data Algorithm

IoT Gateway for Raw Cloud Edge

Data Collection Feature Extraction Computing Computing
Vehicular Network

Collected Sensor Fusion Data

Notifications to Driver, Fleet
Manager, and Service Centre

Real Time Processing

Preprocessing

Historical Data Analysis

Onboard Remote
Fault Diagnosis Monitoring Monitoring Scalable Infrastructure
Repair & ML/DL Algorithm
Maintenance Integrated VHM Data Processing &
Storage

Fig. 8. VHM and fault diagnosis framework

49
 Key Themes
Managing massive, real- Data breaches,
time data streams; ensuring unauthorized access,
interoperability malicious attacks

Advanced data fusion, Secure communication

standardization, cloud/edge protocols, encryption, real-
computing, AI-driven signal time threat detection, AI-
integrity monitoring Data Integration Cybersecurity driven anomaly detection

Accuracy and reliability of Achieving self-awareness;

predictive models; handling real-world
integration with vehicle uncertainties
systems

Machine learning, ML models, self-learning

dynamic Bayesian networks, Prognostics & Automated Fault
systems, integration with
cloud-based analytics, edge Health Management Diagnosis
IoT and edge computing
computing

Complexity of vehicle Network congestion,

systems; information environmental challenges,
overload data reliability

Multimodal interfaces Human-Machine Vehicle to Vehicle Robust V2V protocols,

(voice, touch, visual), adaptive Interface Communication integration with health
HMIs, AR for diagnostics monitoring systems, enhanced
collision avoidance
algorithms

Resilient wireless Harsh environmental

protocols, rugged sensors, conditions; high data
integration with edge AI for throughput and low latency
predictive maintenance requirements
Wireless Monitoring

Legends Challenges Future Directions

Fig. 9. Challenges and future research indications for VHM methods

50
Declaration of interests

☒ The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.

☐ The authors declare the following financial interests/personal relationships which may be
considered as potential competing interests:

Corresponding author:

Prof. Ts. Dr. Md Mustafizur Rahman

Date: 23 March 2025

51