轮式拖拉机电机驱动导航转向系统参数辨识方法

Parameter Identification Method of Wheeled Tractor Motor Drive Navigation Steering System

摘要:

农业机械导航系统应用日益广泛，但在地形起伏大、作业环境恶劣及作物类型复杂的工况，如热带区、丘陵地区，会显著降低导航系统的路径跟踪精度与鲁棒性。转向系统是农机导航系统的核心部件，精准的转向控制是实现高精度路径跟踪的前提。获取准确的转向系统数学模型可有效降低转向控制算法的设计难度与复杂性。然而，由于转向系统中存在动态滞后、非线性摩擦、死区效应及环境相关扰动等未建模动态，采用机理分析方法难以建立准确的转向系统数学模型。为此，本文提出了一种基于卡尔曼滤波 (KF) 与带正则项递归最小二乘法 (RRLS) 联合辨识的转向系统建模方法。在 Matlab 环境下搭建辨识平台，分别实现并比较 RLS,KF+RLS 与 KF+RRLS3 种辨识方案的仿真结果，筛选出性能最优的模型为 KF+RRLS。随后开展田间复杂工况验证试验，结果表明：所辨识模型对轮角响应的轨迹预测与实测数据的吻合程度为 95.26%, 关键动态特性被可靠复现，且满足实时性要求。该模型为复杂环境下农机导航系统的路径跟踪控制算法设计提供了可靠的模型基础与工程化支撑。

Abstract:

The application of agricultural machinery navigation system is becoming more and more extensive, but the path tracking accuracy and robustness of the navigation system will be significantly reduced under the working conditions of large terrain, harsh working environment and complex crop types, such as hot areas and hilly areas. Steering system is the core component of agricultural machinery navigation system. Accurate steering control is the premise of achieving high-precision path tracking. Obtaining an accurate mathematical model of the steering system can effectively reduce the design difficulty and complexity of the steering control algorithm. However, due to the unmodeled dynamics such as dynamic lag, nonlinear friction, dead zone effect and environment-related disturbance in the steering system, it was difficult to establish an accurate mathematical model of the steering system by using the mechanism analysis method. Therefore, a steering system modeling method based on Kalman filter (KF) and regularized recursive least squares ( RRLS) joint identification was proposed. The identification platform was built in Matlab environment, and the simulation comparison of RLS, KF + RLS and KF + RRLS identification schemes was realized and compared, respectively. The model with the best performance was selected as the system model identified by KF + RRLS. Subsequently, a field verification test under complex working conditions was carried out. The resulted showed that the trajectory prediction of the wheel angle response of the identified model was 95. 26% consistent with the measured data, and the key dynamic characteristics were reliably reproduced and met the real-time requirements. The model provided a reliable model basis and engineering support for the design of path tracking control algorithm of agricultural machinery navigation system in complex environment.