Computational Improvements in the Multivalued Fuzzy Behavior Control System for Robot Navigation

Multivalued fuzzy behavior control systems for robot navigation have demonstrated more system robustness against sensor and environmental uncertainty and changes in behavioral priorities than univalued fuzzy behavior control systems. However, a major concern for the implementation of multivalued systems is the computational burden associated with the fuzzy inference and deffuzification process. This paper presents a multivalued fuzzy behavior control system for robot navigation using singletons instead of fuzzy set consequences. The proposed implementation with singleton consequences greatly reduces the space required for data storage and speeds up the control computation while providing a simple way of fusing the behavioral commands. Simulation results show that the singleton implementation yields a navigation trajectory that is roughly the same as or even better than that of the original system that uses fuzzy set consequences. Real-time implementation at high speed is possible with this new type of fuzzy model because of the increased computational efficiency.