Department of Secured Smart Electric Vehicle, Kookmin University, Seoul, 02707, Republic of Korea.
School of Electrical Engineering and Department of Secured Smart Electric Vehicle, Kookmin University, Seoul, 02707, Republic of Korea.
PLoS One. 2018 Mar 13;13(3):e0194110. doi: 10.1371/journal.pone.0194110. eCollection 2018.
In this paper, explicit Model Predictive Control(MPC) is employed for automated lane-keeping systems. MPC has been regarded as the key to handle such constrained systems. However, the massive computational complexity of MPC, which employs online optimization, has been a major drawback that limits the range of its target application to relatively small and/or slow problems. Explicit MPC can reduce this computational burden using a multi-parametric quadratic programming technique(mp-QP). The control objective is to derive an optimal front steering wheel angle at each sampling time so that autonomous vehicles travel along desired paths, including straight, circular, and clothoid parts, at high entry speeds. In terms of the design of the proposed controller, a method of choosing weighting matrices in an optimization problem and the range of horizons for path-following control are described through simulations. For the verification of the proposed controller, simulation results obtained using other control methods such as MPC, Linear-Quadratic Regulator(LQR), and driver model are employed, and CarSim, which reflects the features of a vehicle more realistically than MATLAB/Simulink, is used for reliable demonstration.
本文采用显式模型预测控制（MPC）来实现自动驾驶车道保持系统。MPC 已被视为处理此类约束系统的关键。然而，在线优化的 MPC 存在巨大的计算复杂性，这一直是限制其目标应用范围的主要缺点，使其只能应用于相对较小和/或较慢的问题。显式 MPC 可以使用多参数二次规划技术（mp-QP）来降低这种计算负担。控制目标是在每个采样时刻推导出最优的前轮转向角，以使自动驾驶车辆以高入口速度沿期望路径行驶，包括直线、圆形和回旋线部分。就所提出的控制器的设计而言，通过仿真描述了在优化问题中选择加权矩阵的方法和路径跟踪控制的时滞范围。为了验证所提出的控制器，还使用了其他控制方法（如 MPC、线性二次调节器（LQR）和驾驶员模型）获得的仿真结果，并使用比 MATLAB/Simulink 更真实地反映车辆特性的 CarSim 进行可靠的演示。
