Control of Unmanned Bicycle Based on Tensor Product Model Transformation and Hammersley Sampling Method

In this paper, path planning and balance control of unmanned bicycles are investigated. Firstly, the Deep Deterministic Policy Gradient (DDPG) algorithm is used to train the Artificial Potential Field (APF) algorithm to implement path planning in complex maps. Secondly, a dual closed-loop control strategy is adopted for the path tracking and the balance control. In the inner-loop control, the Hammersley sampling (HS) method is used to obtain the hyper-cube network model of the unmanned bicycle, and the balance controller is designed based on the tensor product (TP) model transformation and parallel distributed compensation (PDC) method for stabilization of the unmanned bicycle. The outer-loop control system uses the Ackermann steering geometry model to track the planned path. Finally, the simulation results demonstrate that the path planning algorithm plans the path that satisfies the constraints and the designed balance controller can keep the unmanned bicycle balanced well during trajectory tracking.