Development of a robotic arm-gantry simulator with probabilistic inference based control

Abstract

Robotic arms can perform repetitive tasks with high speed, accuracy, and precision, making them highly suitable for applications that are monotonous or require a high degree of precision. Such applications include industrial applications, human-robot collaboration applications, and experimental setups for model testing. A robust simulation as prototype testing is typically required to identify potential risks of catastrophic damage to the model or the experimental setup. Robot Operating System (ROS) is a widely-used framework for creating robotic applications both in research and in industry due to its easy hardware abstraction, code re-usability, and compatibility with popular open-source libraries. This thesis is focused on a robotic system consisting of a 7-DOF robotic arm ceiling mounted on a two-axis gantry. A simulator is developed for the 9-DOF robotic system in the ROS-Gazebo framework that interfaces with Moveit API for motion planning. A well-known Approximate Inference Control (AICO) algorithm is added to the simulator to extend its functionality to perform optimal trajectory planning. This work also extends the study of the AICO algorithm to non-holonomic mobile robots. Simulation experiments are conducted on both systems to study the behavior, performance, and limitations of AICO.