Fuzzy logic-based planning and reactive control of autonomous mobile robots

Certain important control problems require the regulation of complex systems operating in uncertain environments. In these problems, the controller seeks to attain a number of explicitly stated goals as well as to meet implicit requirements to respond to unexpected events. We present a fuzzy-logic approach for the treatment of this type of control problems based on the combination of purposive behaviors seeking explicit goals with reactive behaviors intended to respond to unforeseen circumstances. This approach is based on the context-dependent activation of special constructs called control structures. Inferential operations performed on these control structures result in the computation of desirability measures, which quantify the utility associated with the activation of each control action from the viewpoint of a specific goal. Desirability measures, corresponding to multiple, possibly conflicting, goals, are then combined to arrive to an integrated desirability distribution. The actual control value sent to the plant is then computed using defuzzification methods. We illustrate our approach with examples from its application to the control of autonomous mobile intelligent devices