Path following control of nonlinear bicycle model using online learning
Seungjoon Lee, Taewan Kim, H. Jin Kim · 2017
In this paper, we present online learning control of a nonlinear bicycle model based on Echo State Network (ESN). We update only the output weights of the network to reduce calculation load so that the control algorithm works online. The advantage of the online learning control is that it does not need a complex modeling process. Also, it continues learning which improves robustness. The algorithm learns the inverse model of the unknown plant with trial and error. And it creates control inputs to follow the desired path. The overall objective is to track the path without any modeling process. The simulation results are presented.