A Fuzzy Logic-Based Robot Controller

Abstract

Conventional robot controllers rely on an exact inverse kinematic model to convert user-specified Cartesian trajectory commands to joint set points. These inverse kinematic equations are robot specific and form a set of highly coupled nonlinear equations that pose a considerable computational bottleneck. It is not uncommon for the host CPU to dedicate as much as 80% of its computational resources to solve the inverse kinematic equations alone. Further, the strictly mathematical nature of the inverse kinematic equations results in kinematic singularities in which the robot is uncontrollable. These drawbacks have resulted in current robot controllers being limited in flexibility with respect to adaptive control, kinematic redundancy, and obstacle avoidance. This observation provided the impetus to investigate the applicability of fuzzy logic to resolve the computational bottleneck of the inverse kinematic module. In the following, an approach is presented based on fuzzy logic, which replaces the computationally intensive inverse kinematic equations and lays the foundation for a robot controller with a simpler architecture. A hardware implementation of a fuzzy logicbased robot controller is described in which simple fuzzy rules replace the complex inverse kinematic equations. The implemented fuzzy robot controller was tested on a planar four-degree-of-freedom lab robot. Hardware test results of the lab robot tracking arbitrary trajectories are discussed.