Echo state network based dual adaptive control for trajectory tracking of Wheeled Mobile Robot

The trajectory tracking problem of Wheeled Mobile Robots (WMRs) is studied considering both the kinematic and dynamic models. To deal with the unknown dynamic equations and external disturbances, a dual adaptive controller is proposed based on the Echo State Networks (ESNs), which are recurrent neural networks with dynamic reservoirs. The unknown nonlinear dynamic functions are approximated by the ESN and the output weights of ESN leading from the internal neurons to the output neurons are adjusted by conventional Kalman filter. The trajectory tracking dual adaptive controller is online calculated to optimize an explicit-type suboptimal innovation based cost function. The effectiveness of proposed methods is finally verified by simulations of WMR trajectory tracking.