Fuzzy Membership Scaling Mechanisms for Mobile Robot Behaviours

Fuzzy behaviours are commonly used in reactive mobile robot navigation strategies, where sensory information is either uncertain or incomplete. However, the complexity of such controllers usually grow exponentially with the number of fuzzy input partitions and rules in the rule base. Furthermore, attempts to reduce the number of input partitions will typically erode the performance of the controllers. This work investigates several membership function scaling mechanisms as an avenue for improving the performance of fuzzy behaviours based on minimal rule base controllers. The configurations are based on the closely-related concepts of linguistic hedges and non-linear scaling. The scaling parameters for the goal seeking and obstacle avoidance behaviours are tuned in simulation via a genetic algorithm optimisation process. The results show that the controller configuration based on input membership function scaling consistently outperforms simple fuzzy logic controllers with the same number of fuzzy input partitions and rules.