Multi-Layered Fuzzy Behavior Fusion for Reactive Control of Autonomous Robots

The success of fuzzy control systems is owed in large part to the technology’s ability to convert qualitative linguistic descriptions (often provided by human experts) into nonlinear mathematical functions. By bridging the gap between human expert knowledge and the world of embedded digital systems, fuzzy logic has found use in many consumer products. Most commercial fuzzy con-trol implementations feature a single layer of fuzzy inference between two or three inputs and one or two outputs. For autonomous robots, however, the numbers of inputs and outputs are usually larger, and the desired control behaviors are much more complex. This paper presents some important issues regarding the scalability of fuzzy control techniques as applied to autonomous robot behavior, and describes the approaches used on robots at NCSU.