Intelligent Agents
By: Ashok Basnet
Agents and Environment
An agent is anything that can be viewed as perceiving its
environment through sensors and acting upon that environment
through actuators.
We use the term percept to refer to the agent's perceptual inputs at
any given instant.
The agent's behavior is described by the agent function that maps
any given percept sequence to an action.
The agent function for an artificial agent can be implemented by
agent program.
The agent function is abstract mathematical description; the agent
program is a concrete implementation, running within some physical
system.
Agents and Environment
Agent
Function
Agents Program
The Nature of Environments
In our discussion of the rationality of the simple vacuum-
cleaner agent, we had to specify the performance measure, the
environment and the agent's actuators and sensors.
We group all these under the heading of the task environment
and also called as PEAS (Performance, Environment, Actuators,
Sensors).
The vacuum world was a simple example; let us consider a
more complex problem; an automated taxi driver.
Specifying the task environments
The Nature of Environments
First, what is the performance measure to which we would like our
automated driver to aspire? (Correct destination, minimizing fuel)
What is the driving environment that the taxi will face? (Roads,
other traffic, pedestrians)
The actuators for an automated taxi include those available to a
human driver: control over the engine through the accelerator and
control over steering and braking.
The basic sensors for the taxi will include one or more controllable
video cameras so that it can see the road; it might augment these
with infrared or sonar sensors to detect distances to other cars
and obstacles.
The Structure of Agents
Four basic agents:
a. Simple Reflex Agents
b. Model-based reflex agents
c. Goal-based agents
d. Utility-based agents
Each kind of agent program combines particular components in
particular ways to generate actions.
If we convert each of these agents into learning agents then it
can improve the performance of their components so as to
generate better actions.
Simple Reflex Agents
Simple Reflex Agents
This agent selects actions based on the agents current perception or the
world and not based on past perceptions.
For example if a mars lander found a rock in a specific place it needed to
collect then it would collect it, if it was a simple reflex agent then if it found
the same rock in a different place it would still pick it up as it doesn't take
into account that it already picked it up.
This is useful for when a quick automated response is needed, humans
have a very similar reaction to fire for example, our brain pulls our hand
away without thinking about any possibility that there could be danger in
the path of your arm. We call these reflex actions.
This kind of connection where only one possibility is acted upon is called a
condition action rule, written as:
if hand is in fire then pull away hand
Model-based Reflex Agents
Model-based Reflex Agents
Model-based Reflex Agents
Model-based reflex agents are made to deal with partial accessibility; they
do this by keeping track of the part of the world it can see now.
It does this by keeping an internal state that depends on what it has seen
before so it holds information on the unobserved aspects of the current
state.
This time out mars Lander after picking up its first sample, it stores this in
the internal state of the world around it so when it come across the
second same sample it passes it by and saves space for other samples.
While reading this you are keeping track of where you have got to
somewhere internally in your brain just in case you lose your place.
Goal-based Agents
In life, in order to get things done we set goals for us to achieve, this pushes us
to make the right decisions when we need to.
A simple example would be the shopping list; our goal is to pick up everything
on that list. This makes it easier to decide if you need to choose between milk
and orange juice because you can only afford one. As milk is a goal on our
shopping list and the orange juice is not we chose the milk.
So in an intelligent agent having a set of goals with desirable situations are
needed. The agent can use these goals with a set of actions and their predicted
outcomes to see which action(s) achieve our goal(s).
Achieving the goals can take 1 action or many actions. Search and planning are
two subfields in AI devoted to finding sequences of actions to achieve an
agents goals.
Unlike the previous reflex agents before acting this agent reviews many actions
and chooses the one which come closest to achieving its goals, whereas the
reflex agents just have an automated response for certain situations.
Utility-based Agents
Utility-based Agents
Just having goals isn’t good enough because often we may have several actions
which all satisfy our goal so we need some way of working out the most efficient
one.
A utility function maps each state after each action to a real number
representing how efficiently each action achieves the goal.
This is useful when we either have many actions all solving the same goal or
when we have many goals that can be satisfied and we need to choose an
action to perform.
For example let’s show our mars Lander on the surface of mars with an obstacle
in its way. In a goal based agent it is uncertain which path will be taken by the
agent and some are clearly not as efficient as others but in a utility based agent
the best path will have the best output from the utility function and that path
will be chosen.
Learning Agents
Learning Agents
When we expand our environments we get a larger and larger amount of
tasks, eventually we are going to have a very large number of actions to
pre-define.
Another way of going about creating an agent is to get it to learn new
actions as it goes about its business, this still requires some initial
knowledge but cuts down on the programming greatly.
This will allow the agent to work in environments that are unknown.
The learning element is responsible for improvements this can make a
change to any of the knowledge components in the agents.
One way of learning is to observe pairs of successive states in the
percept sequence; from this the agent can learn how the world evolves.
Learning Agents
Learning element: It is responsible for making improvements by learning
from the environment
Critic: The learning element takes feedback from critics which describes
how well the agent is doing with respect to a fixed performance standard.
The learning agent gains feedback from the critic on how well the agent is
doing and determines how the performance element should be modified if
at all to improve the agent.
Performance element: It is responsible for selecting external action.
Problem Generator: This component is responsible for suggesting actions
that will lead to new and informative experiences.
How the components of agent programs work?
Thank you
