Optimal membership functions for multi-modal control

In order to manage the rapidly growing complexity associated with many modern control applications, multi-modal control has emerged as a viable option. In this paper, we assume that an ensemble of individual controllers or modes have been designed, and concentrate on the problem of combining these controllers in an optimal manner. In particular, we use variational arguments for shaping parameterized membership functions in order to minimize a given performance cost, thus resulting in a systematic method for optimal mode fusion in a general setting. This view complements the standard fuzzy logic control approach, where the modes are combined according to some given (and possibly fine-tuned) membership functions. Moreover, the new controller fusion methodology is transitioned onto a real robotic platform navigating in an unknown environment to compare the performance of the proposed method with the standard approach.