Collision avoidance for collaborative robotics

Abstract

In this thesis we visit the problem of real-time collision avoidance for robotic manipulators in unstructured and dynamic environments. The main objective will be to implement a human-robot collision avoidance algorithm for a robotic cell that utilizes an industrial robotic manipulator where the human coworker and the robot share the same working area. For this purpose the pioneering work of Khatib in the arti cial potential eld method was taken as the basis to our work. Thus, an implementation of two di erent collision avoidance controllers is addressed. The rst of which is based on kinematics, while the other is based on force control that take into consideration robot's dynamics. For developing the force controller, and for achieving real time performance of the virtual-reality simulations, we had to implement a light weight numerical method for computing robot's dynamics. Several e cient algorithms for calculating robot dynamics were deduced. Results indicate that the proposed methods compare favourably with state-of-the-art methods. Throughout this work MATLAB® was opted as the tool for implementing the algorithms, while the real-time virtual-reality simulations were carried out using the Virtual Experimentation Platform (V-REP). Using these tools several controllers, algorithms, techniques and simulations were applied and the results achieved were discussed. We conclude this study by identifying some of the issues associated with the arti cial potential eld method