基于多目标优化的策略型自适应农机路径跟踪控制方法

Adaptive Path Tracking Predictive Control Method for Agricultural Machinery Based on Strategy-based Multi-objective Optimization

摘要:

为提升农机路径跟踪平滑度和精度，降低环境噪声、传感器噪声等外部干扰，提出一种基于多目标优化的策略型自适应农机路径跟踪控制方法。以综合误差最小为目标，建立农机运动学模型及误差模型，采用拉丁超立方采样、策略型早停机制和适应度记忆对北极海鹦算法进行优化，利用优化后北极海鹦算法对模型预测算法的元参数进行自适应调整；以减少外部干扰并提升路径平滑程度为目标，建立农机状态多目标优化函数，引入多目标辅助优化算法，并与模型预测算法代价函数结合，对农机控制量进行求解。在此基础上引入事件触发的热启动技术，利用历史数据缩短模型预测控制优化时间。仿真试验结果表明，当农机作业速度为1.0 m/s时，最大绝对误差为0.06 m，平均误差为0.02 m。相较于原预测算法，单次运行时间仅增加0.007 s，路径平滑度平均提升83%。实地试验结果表明，当速度为0.5、1.0、1.5 m/s时，优化后算法平均误差相较于原始模型预测算法分别提升33%、35%、38%，路径平滑程度分别提升40%、51%、10%。

Abstract:

Aiming to enhance the path tracking capability of agricultural machinery in complex environments, an adaptive predictive control method was proposed-based on multi-objective optimization. The goal was to reduce external disturbances and improve path smoothness. Firstly, a kinematic model and error model of the machinery were developed, and its dynamic behavior under working conditions was analyzed. The arctic parrot algorithm was introduced, with a comprehensive error objective function designed for path tracking. By combining real-time feedback, AP adjusted model predictive control (MPC) parameters for better accuracy. Next, a multi-objective optimization algorithm was integrated with the MPC cost function to improve tracking accuracy, smoothness, and stability. To address delays caused by increased controller dimensionality, Latin hypercube sampling was used for efficient population initialization, reducing computational load. An early stopping mechanism and fitness memory were applied to accelerate the optimization process by halting iterations once a fitness threshold reached. Additionally, a warm start technique-based on historical data was introduced to shorten optimization time, enabling faster application to new tasks. Simulation results at 1.0 m/s showed a lateral maximum absolute error of only 0.06 m, with an average error of 0.02 m, while running time remained comparable to traditional MPC algorithms. Path smoothness was improved by 83%, indicating enhanced stability. In field tests, the algorithm outperformed traditional MPC with error reductions of 33%, 35%, and 38% at speeds of 0.5 m/s, 1.0 m/s, and 1.5 m/s, respectively. Path smoothness was increased by 40%, 51%, and 10%. These results validated the effectiveness of this method in practical applications, ensuring stable performance across complex scenarios and reducing path deviations due to external factors.