A new approach for designing TSK fuzzy systems from input-output data

We propose a new approach for determining fuzzy membership functions for system inputs and outputs and for defining the rule base systematically from input-output pairs. This approach treats a human-operated system as a black box and does not require knowledge of its underlying mathematical model. We provide an upper bound on the error resulting from approximating the input-output data using a fuzzy system. To demonstrate the ease and power of the new approach, we apply it to the benchmark problem of backing up a truck into a loading dock. The results show that the fuzzy controller can be chosen as a compromise between the conflicting requirements of tracking accuracy and simplicity of implementation.