In this paper, a robust control architecture is proposed for lane-keeping and obstacle avoidance of autonomous ground vehicles. A two-level hierarchical controller is used to separate the planning and tracking problems. At the higher-level, we solve a nonlinear model predictive control (MPC) problem with an oversimplified point-mass model. The desired trajectories are generated and fed into the lower-level controller, where a force-input nonlinear bicycle model is considered to set up the tracking control law. Moreover, at each time step, a linearized bicycle model is derived and implemented to reduce the real-time computational complexity. Based on the above profile, a discrete-time integral sliding MPC (DISMPC) technique is used to improve the system robustness. By introducing an additional sliding control term into the feedback control law, the system trajectories can be maintained within a quasi-sliding band. In this case, it becomes necessary to take into account the system dynamics induced by the sliding control. Namely, the state and the input constraints of the MPC problem at each level need to be tightened. This helps to guarantee the feasibility of the original constrained problem in the presence of disturbances. Simulations have been carried out to verify the effectiveness of the proposed controller. The results show that the controller is able to simultaneously achieve lane-keeping and obstacle avoidance with uncertain friction coefficients.
UAV Formation Flight Based on Nonlinear Model Predictive Control
