Authors

Abstract

For the automatic control of soybean harvester header height, this study uses Area array LiDAR for header height detection. An improved quartile range algorithm is used to dynamically remove outliers under crop residue interference. Linear, quadratic, and cubic nonlinear terrain fitting models are established based on the surface undulation characteristics of soybean fields. The Huber loss function is introduced to enhance the robustness of parameter estimation. The balance between model complexity and fitting goodness is quantified using Bayesian information criterion (BIC), and the model intercept term with the smallest BIC value is selected as the terrain reference height. Aiming at the hysteresis characteristics of valve controlled asymmetric hydraulic cylinders, a telescopic dual-mode transfer function model is established, and a Bang Bang switch lead compensation strategy with position threshold is proposed. By predicting the trend of terrain changes, the electromagnetic directional valve is triggered in advance when the height error of the header exceeds the set threshold, effectively reducing the system response delay. Field comparative experiments have shown that at a working speed of 1m/s, the automatic control mode significantly improves the uniformity of cutting compared to the manual mode. When the cutting threshold is set to 20, 25, and 30mm, the coefficient of variation of cutting height is reduced by 2.13%, 1.71%, and 0.55%, respectively. Moreover, the automatic mode maintains a gentle distribution characteristic within the threshold range of 15-35mm, verifying the strong robustness and control accuracy advantages of the designed system in complex farmland environments.

Abstract in Chinese

针对大豆收割机割台高度自动控制，本研究采用面阵激光雷达进行割台高度检测，利用改进四分位距算法实现作物残茬干扰下的离群点动态剔除，并依据大豆田地表起伏特征建立线性、二次及三次非线性地形拟合模型，通过引入Huber损失函数增强参数估计鲁棒性，结合贝叶斯信息准则（BIC）量化模型复杂度与拟合优度的平衡关系，择优选取BIC值最小的模型截距项作为地形基准高度。针对阀控非对称液压缸的迟滞特性，建立伸缩双模态传递函数模型，提出带位置阈值的Bang-Bang开关超前补偿策略：通过预判地形突变趋势，在割台高度误差超过设定阈值时提前触发电磁换向阀动作，有效降低了系统响应延迟。田间对比试验表明，在1m/s作业速度下，自动控制模式较手动模式显著改善割茬均匀性，当割茬阈值设置为20、25及30mm时，割茬高度变异系数分别降低了2.13%、1.71%及0.55%，且自动模式在15-35 mm阈值范围内始终维持平缓分布特征，验证了所设计系统在复杂农田环境中的强鲁棒性与控制精度优势。