线性时变模型预测控制器提高农业车辆导航路径自动跟踪精度

发布时间：2018-04-15 06:06

本文选题：车辆 + 控制　；参考：《农业工程学报》2017年13期

【摘要】：为提高农业车辆导航路径自动跟踪精度,提出一种基于线性时变模型预测控制的路径跟踪方法。该方法将农业车辆非线性运动学模型线性化和离散化处理,作为控制器预测方程;建立以系统控制增量为状态量的目标函数,为防止无可行解,引入松弛因子;设计系统控制量、控制增量和状态量约束条件,并将目标函数求解转为带约束的二次规划问题;采用内点法进行求解,将求得的控制输入增量第一个元素作用于系统;重复以上过程,实现优化控制。基于Matlab/Simulink平台进行了模型预测控制器设计,并分别进行了导航坐标系下的直线和圆形路径跟踪试验。结果表明,所设计的控制器能够实现直线路径的完全跟踪(误差始终为0);跟踪圆形路径时,1 m/s时的横向平均跟踪误差为7.5 cm,3 m/s时的横向平均跟踪误差为10 cm;整个跟踪过程,前轮转角始终被限定在约束范围内。不同控制器参数下的仿真结果表明,增大预测时域和控制周期能够减小跟踪误差和前轮转角变化幅度,控制时域的变化对控制器路径跟踪响应速度影响较小。同时基于设计的模型预测控制器进行了场地试验。结果表明,试验小车以1m/s的速度跟踪直线路径时,横向最大误差均值为1.622 cm,横向平均误差均值为0.865 cm;跟踪圆形路径时,当行走速度低于1 m/s时,横向最大误差小于10 cm。
[Abstract]:In order to improve the precision of automatic tracking of agricultural vehicle navigation path, a path tracking method based on linear time-varying model predictive control is proposed.In this method, the nonlinear kinematic model of agricultural vehicle is linearized and discretized as the predictive equation of the controller, and the objective function of the system control increment as the state quantity is established. In order to prevent the infeasible solution, the relaxation factor is introduced.The system control quantity, control increment and state constraint conditions are designed, and the objective function is solved into a quadratic programming problem with constraints, and the first element of the obtained control input increment is applied to the system by using the interior point method.Repeat the above process to achieve optimal control.The model predictive controller is designed based on Matlab/Simulink platform, and the linear and circular path tracking experiments in navigation coordinate system are carried out respectively.The results show that the designed controller can completely track the linear path (the error is always zero), and the transverse average tracking error is 10 cm when the tracking error is 7.5 cm ~ (-3) m / s, and the tracking error is 10 cm when the tracking error is 1 m / s.The front wheel angle is always limited to the bound range.The simulation results under different controller parameters show that increasing the predictive time domain and control period can reduce the tracking error and the change amplitude of the front wheel angle, and the change of the control time domain has little effect on the path tracking response speed of the controller.At the same time, the field test of the model predictive controller based on the design is carried out.The results show that the mean value of the transverse maximum error is 1.622 cm and the average lateral average error is 0.865 cm when the speed of the test car is 1m/s, and the transverse maximum error is less than 10 cm when the walking speed is less than 1 m / s.
【作者单位】：山东农业大学机械与电子工程学院;山东省园艺机械与装备重点实验室;山东农业大学信息科学与工程学院;
【基金】：山东省现代农业技术体系创新团队岗位专家资助项目(SDAIT-10-022-10) 山东农业大学博士后经费(76567);山东农业大学青年科技创新基金(24158)
【分类号】：S22;TP273

【参考文献】