tud-cor / jackal_active_inference_versus_kalman_filter

ROS package for controlling the existing Jackal simulation in Gazebo (built by Clearpath Robotics Inc.) and using the simulation data to run an Active Inference algorithm and Kalman filter

Design of an Active Inference and Kalman filter applied to the Jackal robot simulation in Gazebo 7 using ROS Kinetic

This ROS package contains the following folders:

• launch: includes the launch file needed to run all ROS nodes at once, including the simulation control, Gazebo simulation with correct arguments, data processing, Active Inference algorithm and Kalman filter
• scripts: includes all Python scripts needed to run the simulation control, data processing, Active Inference algorithm and Kalman filter ROS nodes
• msg: includes all custom ROS message definitions used for the data processing and filtering nodes
• worlds: defines the Gazebo simulation world and the physics engine with corresponding necessary properties to correctly simulate the Jackal robot
• config: includes the yaml file needed to adjust the publish rate of the /joint_states topic, needed to have approximately equal control input and system output update frequencies
• matlab: includes MATLAB files used to analyse the simulation and filter results
• bagfiles: includes recorded ROS topics used in post-simulation analysis in MATLAB
• doc: includes the project report and documentation to install all required software and reproduce the obtained simulation results. See below

This ROS package contains the following files:

• CMakeLists.txt: build script for the ROS package
• package.xml: contains meta-data and dependencies of the ROS package
• LICENSE: all contributing software in this package is licensed under Apache License 2.0

Documentation

report

This document is the result of the project during which this package was created. It explains the design, implementation and validation steps of the software architecture and the implemented algorithms. It furthermore provides theoretical conclusions regarding the Active Inference algorithm as well as the Kalman filter.

This document is useful for people looking to understand the software architecture as well as the theory behind the implemented algorithms.

installation_manual

This documents contains the instructions to install the software on an Ubuntu 16.04 operating system. The installation manual of the following software parts is included:

• ROS Kinetic
• Gazebo 7
• Gazebo 9
• Jackal packages for simulation in Gazebo
• MATLAB R2019a

Note that this manual is partly finished yet! Some installation details still need to be verified.

system_overview

This document provides an overview of all tunable parameters in the system, sorted by system part:

• Simulation control
• Gazebo simulation
• Data processing
• Kalman filter
• Active Inference algorithm

This document can especially be useful when trying to improve the simulation results.

Acknowledgements

The jackal_gazebo package is used to start the Gazebo simulator and spawn the Jackal robot. The repository can be found here.