Design and implementation of an adaptive critic-based neuro-fuzzy controller on an unmanned bicycle

TL;DR

This paper presents an adaptive critic-based neuro-fuzzy controller for an unmanned bicycle, utilizing reinforcement learning and feedback to improve control performance through online tuning.

Contribution

It introduces a novel neuro-fuzzy control system that employs critic-based reinforcement learning for real-time adaptation on an unmanned bicycle.

Findings

Improved transient response in control performance

Enhanced robustness to model uncertainties

Fast online learning capability

Abstract

Fuzzy critic-based learning forms a reinforcement learning method based on dynamic programming. In this paper, an adaptive critic-based neuro-fuzzy system is presented for an unmanned bicycle. The only information available for the critic agent is the system feedback which is interpreted as the last action performed by the controller in the previous state. The signal produced by the critic 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 the proposed controller. The results demonstrate superior performance of the developed controller in terms of improved transient response, robustness to model uncertainty and fast online learning.