Fuzzy critic-based learning forms a reinforcement learning method based ondynamic programming. In this paper, an adaptive critic-based neuro-fuzzy systemis presented for an unmanned bicycle. The only information available for thecritic agent is the system feedback which is interpreted as the last actionperformed by the controller in the previous state. The signal produced by thecritic agent is used along with the error back propagation to tune (online)conclusion parts of the fuzzy inference rules of the adaptive controller.Simulations and experiments are conducted to evaluate the performance of theproposed controller. The results demonstrate superior performance of thedeveloped controller in terms of improved transient response, robustness tomodel uncertainty and fast online learning.
展开▼
