Adaptive navigation of an omni-drive autonomous mobile robot in unstructured dynamic environments

One of the challenges of Autonomous Systems navigating in real world is to deal with the large amounts of uncertainties which are inherent in such environment while maintaining stability. Higher order Fuzzy Logic Systems (FLS), such as Interval Type-2 Fuzzy Logic Systems (IT2FLS), that use type-2 fuzzy sets, can model and handle such uncertainties, and give good performances that outperform their Type-1 counterparts. However, the complexity and computational time of type-reduction process which is strongly related to Membership Functions (MFs) structure and the number of fuzzy rules limit their applications to simple cases in real-time. Artificial Potential Field (APF) approach due to its elegant mathematical analysis, simplicity and its possibility to take into account the dynamic of the system is widely used for autonomous mobile robots navigation. However, the potential field introduced exhibits local minima other than at the goal position of the robot. In this paper, a new real-time navigation approach where we combine the APF and IT2FL approaches is developed and implemented for an omnidrive mobile robot navigating in dynamic unstructured environments. The novelty of the approach is the association of IT2FL to APF and the way in which the two approaches are hybridized (rule base size reduction). The experiments carried out on an omnidrive mobile robot named Robotino show the effectiveness of our approach.