Enhanced Fuzzy Algorithms for Robot Control with Collision Avoidance

This thesis deals with fuzzy algorithms for robot control with collision avoidance. The work builds on and extends the PhD thesis of V. Khachatouri, who introduced a new framework for the incorporation of human expertise into robots with fuzzy algorithms. He suggested a new iterative method for the solution of the inverse kinematic problem of redundant robots and a control architecture with enables the combination of different elementary behaviors. For a three-link planar manipulator he derived the algorithms for the inverse kinematics with local collision avoidance. The present thesis contains the following innovations: 1. The collision avoidance strategy is combined with a strategy for joint limit avoidance. 2. The controller architecture is revised and the rule base is clearly arranged and easy to survey. 3. Some important features are added to the existing capabilities, e.g., grasping around an obstacle. 4. The two-dimensional distance calculation needed for the three-link planar robot is generalized to the three-dimensional case. 5. First investigations in the field of three-dimensional collision avoidance by the above mentioned algorithms are done by adding a fourth degree of freedom to the three-link planar robot kinematic. Furthermore, in order to verify these theoretical results the visualization tool RoboVista from INFA is revised to become a comfortable testing platform for robot control algorithms. This tool offers three-dimensional visualization of free definable robot kinematics and various obstacles and is designed to substitute a former two-dimensional visualization also developed by INFA. Finally, to get away from the simulation level, a small three-link manipulator is built, and the performance of the proposed algorithms is tested in real-time.