农业机械自动导航技术研究进展

Review of research on automatic guidance of agricultural vehicles

农业机械自动导航是精准农业技术体系中的一项核心关键技术,广泛应用于耕作、播种、施肥、喷药、收获等农业生产过程。农机位置测量方法、农机模型与导航路径跟踪控制方法是农业机械自动导航技术的研究重点,受到国内外科研人员的广泛关注。农机位置测量主要有相对测量和绝对测量二类方法,前者以基于机器视觉的测量方法为代表,主要利用图像处理技术识别作物行,进而确定导航基准线,实现农机与作物的相对位置与航向信息的测量;后者则以基于全球导航卫星系统的测量方法为代表,利用卫星定位技术实现农机位置的高精度测量,在农业生产中应用最为广泛;而面对复杂的田间环境变化,在位置测量中应用多传感器数据融合技术通常可以得到更好的测量结果。导航路径跟踪控制通常以农机运动学模型或动力学模型为核心,多采用最优控制、最优估计、自适应控制、人工神经网络、模糊控制、鲁棒控制等现代控制理论与方法;而无模型控制方法则可以避免建模不准确或者模型参数剧烈变化对农机路径跟踪控制性能所产生的负面影响。该文从上述2个方面综述分析了农业机械自动导航技术的研究现状及存在的问题,并对未来农机导航技术的发展做出了展望,指出采用卫星导航技术,开展农机地头自动转向控制、障碍物探测及主动避障、多机协同导航等高级导航技术研究,以及引入先进的物联网技术,是现代农机自动导航技术发展的主要趋势。

The automatic guidance of agriculture vehicles is a key technology in precision agriculture and widely used in agriculture production. A review of the recent research in agriculture vehicle automatic guidance is presented in this paper, focusing on the position measurement, agriculture models and path tracking. And some forecasts are made on the trends of the agriculture vehicle automatic guidance. Generally, a modern agriculture vehicle automatic guidance system consists of 4 units： A detecting unit that measures the position and orientation of the vehicle; a control unit, as the core of the guidance system, which makes the plan of the path and carries out the path tracking; an executing unit that makes the turn of the wheels according to the command of the control unit; and a monitoring unit, or a field computer as it is called generally, which works as the interface between human and machine. There are 2 main problems to be solved in the agriculture vehicle guidance system. The first one is the measurement of the agriculture vehicle’s working conditions, such as its position, heading, speed and wheel angle, among which the most important is the position measurement. There are 2 kinds of position measurement methods： One is the relative method, such as measuring the vehicle’s position relative to a guidance baseline based on machine vision; the other is the absolute method, such as measuring the vehicle’s absolute position on the earth based on the Global Navigation Satellite System. As the agriculture vehicle automatic guidance system is working in the field, the complicated and non-structured environment makes none of the measurement methods working well all the time. So the multi-sensor data fusion is brought into sharp focus by researchers. By combining measuring data from different sensors with some data fusing methods, such as Kalman filter, particle filter, H∞ filter, and intelligent methods, the measurement accuracy is improved. The integrated navigation systems are mainly GPS/INS, GPS/DR and INS/CNS. The second problem is agriculture modeling and path tracking control methods. Most of the path tracking control algorithms use kinematics models. The two-wheel model is the most frequently used model, in which an agriculture vehicle is regarded as a two-wheel vehicle and its pose is described by its geographical coordinates, heading, wheel angle and speed. Dynamics models based on the Newton second law are another kind of model commonly used. As it takes into account of the change of the vehicle’s dynamic characteristics with the external environment and the farm implements, it makes the control algorithms more robust. Besides the control methods based on models, researchers have developed some kind of algorithm without a model. The PID is the most useful control strategy. Another one is the pure pursuit method which simulates the driving behavior of human and has foresight. Nowadays, the agriculture vehicle automatic navigation technologies have widely used in the agriculture production, but many problems still need to be studied further. The advanced navigation technologies are worth studying, such as the headland turning control, obstacle detecting and active collision avoidance, and cooperative navigation by multi vehicles. And the agriculture vehicle internet of things is another interesting research area.