Authors

Abstract

This paper presents a multi-objective optimization approach for unmanned agricultural vehicles operating in complex farmland environments. To overcome the limitations of traditional Model Predictive Control (MPC) and heuristic algorithms, a Dung Beetle Optimization-based MPC (D-MPC) multi-objective optimization method is proposed. Specifically, a kinematic model of the unmanned agricultural vehicle is established, incorporating the operational characteristics of complex farmland conditions. The Dung Beetle Optimization (DBO) algorithm is integrated into the MPC framework to enhance performance by leveraging the population-based search behavior of dung beetles. This integration improves both control accuracy and computational efficiency by dynamically adjusting control inputs based on real-time motion predictions, enabling more precise trajectory optimization. Experimental validation is conducted through a dual-verification approach, including both simulation and real-vehicle tests. The results indicate that, compared with conventional control methods, the proposed approach improves trajectory tracking accuracy by approximately 50% and 75% in two representative simulation scenarios, while increasing the battery State of Charge (SOC) by 0.1% and 0.12%, respectively. In real-vehicle experiments, trajectory tracking accuracy is improved by 70%, and SOC is increased by 0.015%.

Abstract in Chinese

本文研究了无人农机在复杂农田场景下的多目标优化问题。针对传统模型预测控制（MPC）和启发式算法容易陷入局部最优或收敛速度慢的问题，提出了一种蜣螂算法优化MPC（D-MPC）的多目标优化算法。首先，针对复杂农田作业的特点，建立了无人农机的运动学模型。其次，采用蜣螂优化算法（DBO）对传统MPC算法进行优化。利用蜣螂种群的搜索特性，显著提高了控制器的控制精度和计算速度。该方法根据实时运动预测调整控制输入，实现对未来状态的预测和优化。最后，设计了一个包括仿真和实车试验的双重验证试验方案。实验结果表明，与传统的控制方法相比，在两种典型的仿真场景下，该方法的轨迹跟踪精度分别提高了50%和75%，而电池的荷电状态（SOC）分别提高了0.1%和0.12%。在实车试验中，轨迹跟踪精度提高了70%，SOC提高了0.015%。