Nonlinear control of static systems with unsupervised learning ofthe initial conditions

摘要

We propose an algorithm for adaptive control of nonlinearmultiple-input, multiple-output (MIMO) static systems. The advantage ofthe proposed method lies in its easy on-line implementation and reliableperformance that is due to an intelligent algorithm for handling theinitial conditions. We represent the nonlinear system with a piecewiselinear model-a variable matrix of first order changes (i.e. the Jacobianmatrix). In order to achieve global optimality for different initialconditions, an enhanced method involving on-line learning a fuzzyrule-base for determining a set of “good” initial conditionsis developed. The knowledge in the rule-base (so as to determine a“good” initial input change) is extracted by on-lineanalyzing the relationship between the error signal, and the controllerresponse. The algorithm continuously evaluates this information, andupdates the rule-base model by using an unsupervised learning scheme.The rule-base provides the “optimal” estimates for theJacobian and initial control settings anytime the control algorithmstarts from new initial conditions