HIERARCHAL CONTROLLER FOR CHVMP-CHAMP QUADRUPED
ROBOT USING GAZEBO - K. ADITYA ANIRUDH
ROS PACKAGES FOR CHAMP QUADRUPED CONTROLLER.
CHAMP is an open source development framework for building
new quadrupedal robots and developing new control
algorithms. The control framework is based on "Hierarchical
controller for highly dynamic locomotion utilizing pattern
modulation and impedance control : implementation on the
MIT Cheetah robot".
Core Features:
1) Fully Autonomous (using ROS navigation Stack). 2) Setup-assistant to configure newly built robots. 3) Collection of pre-configured
URDFs like Anymal, MIT Mini Cheetah, Boston Dynamic's Spot and LittleDog. 4) Gazebo simulation environment. 5) Compatible with DIY
quadruped projects like SpotMicroAI and OpenQuadruped. 6) Demo Applications like TOWR and chicken head stabilization. 7)
Lightweight C++ header-only library that can run on both SBC and micro-controllers.
Tested on:
1) Ubuntu 16.04 (ROS Kinetic) 2) Ubuntu 18.04 (ROS Melodic)
1. INSTALLATION
1.1 Clone and install all dependencies
sudo apt install -y python-rosdep cd /src git clone --recursive https://github.com/chvmp/champ git clone
https://github.com/chvmp/champ_teleop cd .. rosdep install --from-paths src --ignore-src -r -y
1.2 Build your workspace:
cd catkin_make source /devel/setup.bash
2. QUICK START:
You don't need a physical robot to run the following demos. If you're building a physical robot, you can find out more how to configure
and run a new robot in step 3.
2.1 WALKING DEMO IN RVIZ:
2.1.1 Run the base driver:
roslaunch champ_config bringup.launch rviz:=true
2.1.2 Run the teleop node:
roslaunch champ_teleop teleop.launch
If you want to use a joystick add joy:=true as an argument.
2.2 SLAM DEMO:
2.2.1 Run the Gazebo environment:
roslaunch champ_config gazebo.launch
2.2.2 Run gmapping package and move_base:
roslaunch champ_config slam.launch rviz:=true
To start mapping: Click '2D Nav Goal'. Click and drag at the position you want the robot to go.
Save the map by running:
roscd champ_config/maps rosrun map_server map_saver
2.3 AUTONOMOUS NAVIGATION:
2.3.1 Run the Gazebo environment:
roslaunch champ_config gazebo.launch
2.3.2 Run amcl and move_base:
roslaunch champ_config navigate.launch rviz:=true
To navigate: Click '2D Nav Goal'. Click and drag at the position you want the robot to go.
3. RUNNING YOUR OWN ROBOT:
There are two ways to run CHAMP on a real robot:
Linux Machine: Use this ROS package to calculate the joint angles and send it to a hardware interface to control your actuators. You can
follow these guidelines to create your actuators' interface.
Lightweight Version: Run CHAMP's lightweight version on Teensy series microcontrollers and use it to directly control your actuators.
3.1 GENERATE ROBOT CONFIGURATION
First generate a configuration package using champ_setup_assistant. Follow the instructions in the README to configure your own robot.
The generated package contains:
URDF path to your robot. Joints and Links map to help the controller know the semantics of the robot. Gait parameters. Hardware Drivers.
Navigation parameters (move_base, amcl and gmapping). Microcontroller header files for gait and lightweight robot description. This
only applies to robot builds that use microcontroller to run the quadruped controller.
As a reference, you can check out the collection of robots that have been pre-configured here. In the list are some of the popular
quadruped robots like Anymal, MIT Mini Cheetah, Boston Dynamic's LittleDog, and SpotMicroAI. Feel free to download the configuration
packages in your catkin workspaces 'src' directory to try.
Next, build your workspace so your newly generated package can be found:
cd catkin_make
3.2 BASE CONFIGURATION:
This will run the quadruped controller and all sensor/hardware drivers:
roslaunch bringup.launch
Available Parameters: rviz - Launch together with RVIZ. Default: false lite - Always set this to true if you're using a microcontroller to run
the algorithms. Default false.
Example Usage:
View your newly configured robot:
roslaunch bringup.launch rviz:true
Run real robot with a microcontroller:
roslaunch bringup.launch lite:=true
3.3 CREATING A MAP:
The base driver described in 3.2 must be running to run gmapping and move_base.
Run gmapping package and move_base:
roslaunch slam.launch
To open RVIZ and view the map:
roscd champ_navigation/rviz rviz -d navigate.rviz
To start mapping: Click '2D Nav Goal'. Click and drag at the position you want the robot to go.
Save the map by running:
roslaunch navigate.launch
To open RVIZ and view the map:
roscd champ_navigation/rviz rviz -d navigate.rviz
To navigate: Click '2D Nav Goal'. Click and drag at the position you want the robot to go.
3.5 RUNNING YOUR ROBOT IN GAZEBO
Run Gazebo and the base driver in simulation mode:
roslaunch gazebo.launch
Take note that in order for this to work, the URDF has to be Gazebo compatible and has ros_control capability. The controllers have been
set-up so all you need is to add the transmission of the actuators. You also need to get the physics parameters right like your mass,
inertia, and foot friction. Some useful resources on getting these parameters right:
Inertial Calculation - https://github.com/tu-darmstadt-ros-pkg/hector_models/blob/indigo-
devel/hector_xacro_tools/urdf/inertia_tensors.urdf.xacro
List of Moment of Inertia - https://en.wikipedia.org/wiki/List_of_moments_of_inertia
Gazebo inertial parameters - http://gazebosim.org/tutorials?tut=inertia&cat=build_robot#Overview
You can also check out this pull request as an example.
3.6 SPAWNING MULTIPLE ROBOTS IN GAZEBO
Run Gazebo and default simulation world:
roslaunch champ_gazebo spawn_world.launch
You can also load your own world file by passing your world's path to 'gazebo_world' argument:
roslaunch champ_gazebo spawn_world.launch gazebo_world:=
Spawning a robot:
roslaunch champ_config spawn_robot.launch robot_name:= world_init_x:= world_init_y:=
4. TUNING GAIT PARAMETERS
The gait configuration for your robot can be found in /gait/gait.yaml.
Knee Orientation - How the knees should be bent. You can can configure the robot to follow the following orientation .>> .>< .<< .<>
where dot is the front side of the robot.
Max Linear Velocity X (meters/second) - Robot's maximum forward/reverse speed.
Max Linear Velocity Y (meteres/second) - Robot's maximum speed when moving sideways.
Max Angular Velocity Z (radians/second)- Robot's maximum rotational speed.
Stance Duration (seconds)- How long should each leg spend on the ground while walking. You can set this to default(0.25) if you're not
sure. The higher the stance duration the further the displacement is from the reference point.
Leg Swing Height (meters)- Trajectory height during swing phase.
Leg Stance Height (meters)- Trajectory depth during stance phase.
Robot Walking Height (meters) - Distance from hip to the ground while walking. Take note that setting this parameter too high can get
your robot unstable.
CoM X Translation (meters) - You can use this parameter to move the reference point in the X axis. This is useful when you want to
compensate for the weight if the center of mass is not in the middle of the robot (from front hip to rear hip). For instance, if you find that
the robot is heavier at the back, you'll set a negative value to shift the reference point to the back.
Odometry Scaler - You can use this parameter as a multiplier to the calculated velocities for dead reckoning. This can be useful to
compensate odometry errors on open-loop systems. Normally this value ranges from 1.0 to 1.20.
