Direct-reinforcement-adaptive-learning fuzzy logic control for a class of nonlinear systems

The paper is concerned with the application of reinforcement learning techniques to feedback control of nonlinear systems using adaptive fuzzy logic systems (FLS). Even if a good model of the nonlinear system is known, it is often difficult to formulate a control law. The work in this paper addresses this problem by showing how an adaptive FLS can cope with nonlinearities through reinforcement learning with no preliminary off-line learning phase required. The reinforcement learning rules for finding proper fuzzy rules and tuning membership functions (MFs) do not assume that there is a supervisor to decide whether the current control action is correct. Instead, the FLS is indirectly told about the affect of its control action on the system performance. The learning is performed online based on a binary reinforcement signal from a critic without knowing the nonlinearity appearing in the system, and so is called direct reinforcement adaptive learning (DRAL). The learning algorithm is derived from Lyapunov stability analysis, so that both system tracking stability and error convergence can be guaranteed in the closed-loop system.