A Study on the Optimum Design of Fuzzy Logic Control using Differential Evolution Algorithms for Omnidirectional Mobile Robot Navigation

Goal-oriented mobile robot navigation is significant for exploration, transport, and telerobotics ubiquitously. Among the existing mobile robot configurations, the omnidirectional platforms offer enhanced maneuverability and flexibility in navigation. Also, Fuzzy Logic (FL) is favorable regarding intelligible control rules and robustness to uncertainties among the existing control schemes. However, there exists a gap in the control of omnidirectional robots using FL, especially tackling stagnation and local-minima in goal-oriented navigation. This study presents a methodological framework, first of its kind to the best of our knowledge, combining the benefits of a behavioral FL and Differential Evolution for stagnation-free and goal-oriented omnidirectional robot navigation. Whereas FL considers a behavioral modulation between goal-seeking and obstacle avoidance, relevant classes of Differential Evolution (DE), which comprise exploration, exploitation, and self-adaptation modalities, tackle FL membership functions' search space. Through rigorous computational experiments, we demonstrate that Rank-Based Differential Evolution (RBDE)'s exploitative features show favorable performance in learning and tackling stagnation in the context of a small number of function evaluations (up to 1000) and unknown/partially known environments, both of which are relevant for quick adaptation to dynamic environments. After explaining the reason for learning efficiency and navigation performance in training-testing scenarios, we discuss the potential of rank-based learning algorithms.