Abstract:
Abstract: With the steady progress of China's agricultural modernization, broad demands of agricultural machinery for production are growing extensively. Agricultural unmanned aerial vehicles (UAVs), an important part of agricultural aircraft, have become a new force in the field of China Precision Agricultural Aviation. Used in agricultural areas where ground-based machinery has difficulty in cultivating, agricultural UAVs have shown obvious characteristics and advantages in farming, application and popularization. With the further development of precision agricultural aviation technology, one of the inevitable trends of intelligent development is to realize the autonomous identification of obstacles and real-time obstacle avoidance for agricultural UAVs. The ideal obstacle avoidance system can automatically and promptly avoid all kinds of obstacles in the flight path, avoiding accidents due to operational errors, failures of autonomous flight or other unexpected failures. Thus, it can effectively reduce unnecessary loss of property and casualties. However, the complex operating environment of farmland and the changing types of obstacles both increase the complexity of obstacle avoidance research. When operating during the outbreak of pests, at night or in an environmental with low visibility, many factors will limit the work efficiency and increase the risk of crash if the UAVs mainly rely on pilots to observe and make judgement. Therefore, much progress needs to be made to realize real-time obstacle avoidance of agricultural UAVs. To improve the safety of aviation operations, in this paper, we classified the farmland obstacles in two ways and puts forward the obstacle avoidance zone and related avoidance tactics. The distance between obstacles and the movement of agricultural UAVs was relative. However, the obstacle avoidance zone of its main field of view was fixed and absolute. In execution area (<5 m), the obstacle avoidance motion commands shall be executed to avoid short-range obstacles. In warning region (5-15 m), early warning shall be automatically performed on the medium-long-distance obstacles, and follow-up shall be carried out to prevent unexpected accidents. The obstacles in safe region (>15 m) can be ignored temporarily until they entered the warning region, then it should start the obstacle avoidance pretreatment. Because large, small and medium-sized obstacles had obvious features, they can be totally or partially detected by obstacle avoidance sensors in the safe area. When entering the warning area or execution area, it was relatively easy for drones to avoid them. As for micro and visually-distorting obstacles, it was difficult to identify and evade them even when they got into the warning and execution areas. At present, apart from improper manual operation or sudden failure of the machine, the main reason for most UAV crash accident was obstacle avoidance failure, such as failing to avoid wires, branches and other tiny obstacles. Although the rate of crash accidents caused by tiny obstacles was relatively low, the property loss rate and the casualty rate were extremely high. Therefore, how to avoid tiny obstacles should be the priority of agricultural UAV obstacle avoidance research. Through comparative analysis of domestic and foreign application situation of agricultural UAV obstacle avoidance technology and various obstacle avoidance sensors, in this paper, we analyzed the deficiency of China's agricultural UAV obstacle avoidance problem. With the further development of precision agricultural aviation technology, avoidance system with multi-sensor fusion would become the mainstream of real-time obstacle avoidance system for agricultural UAVs. The combination of visual and non-visual sensors provided a variety of possibilities for the implementation of autonomous spraying and intelligent navigation. Finally, in this paper, we proposed several forecasts to the UAV obstacle avoidance system construction, including development of real-time active obstacle avoidance technology, the agricultural UAV obstacle avoidance process and an obstacle avoidance method to identify the substitutes of tiny obstructions, in order to provide a reference for the orderly development of agricultural UAV's obstacle avoidance technology in China.