Adaptive neural control of wheeled inverted pendulum models

This paper investigate motion control of wheeled inverted pendulum (WIP) models, which have been widely applied for a large class of modern vehicles that can transport human with high safety and work capability. Neural network (NN) has been employed to design adaptive control for the fully actuated tilt and yaw angular motion subsystem using a reference model derived by finite time linear quadratic regulation (LQR) optimization technique. The forward velocity is indirectly “controlled” by the implicit control trajectory, which is then planned by an NN based adaptive generator of implicit control trajectory (AGICT).