An Online Model Predictive Control Framework for Robot Driver Speed Control
An Online Model Predictive Control Framework for Robot Driver Speed Control
复制标题
机器人驾驶员速度控制的在线模型预测控制框架
DOI：
10.1115/dscc2018-8957
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Dimitar Filev
中科院分区：
文献类型：
--
作者：
Jing Wang;Yan Wang;Dimitar Filev
针对机器人驾驶员速度控制问题，提出了一种在线自适应模型预测控制（MPC）方法。该控制结构包含三个基本组成部分：正则化最小二乘法用于识别车辆系统的时变状态空间模型；利用卡尔曼滤波器（KF）识别模型的内部状态；以及基于合理构造的优化目标和约束生成控制输入的模型预测控制器（MPC）。我们将机器人驾驶员速度跟踪问题视为加速度跟踪问题，并将其求解为一个具有时变约束的优化问题，这些约束依赖于车辆的速度和加速度。在FTP城市驱动周期上的仿真结果表明，该方法具有良好的速度跟踪能力，且标定量很小。该控制器在不降低控制性能的前提下，能够适应车辆重量的变化。
In this paper, we present an online adaptive model predictive control (MPC) method for the robot driver speed control problem. The control structure contains three essential components: a regularized least square method to identify a time varying state space model for the vehicle system; a Kalman Filter (KF) to identify the internal states of the model; and a model predictive controller (MPC) to generate the control input based on properly constructed optimization objective and constraints. We cast the robot driver speed tracking problem as an acceleration tracking problem which we solve as an optimization problem with time varying constraints that are dependent on vehicle speed and acceleration. Simulation results on the FTP city drive cycle shows that the proposed approach has good speed tracking capability with minimal calibration efforts. and the controller is adaptive to variation on vehicle weight without degrading control performance.