Abstract:
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.