Path Planning And Control Method For Automatic Driving Of Agricultural Machinery

The transformation and upgrading of the agricultural machinery equipment industry is included in the "Made in China 2025" plan,and the automatic driving of agricultural machinery is an important means to promote agricultural automation and intelligence.Self-driving agricultural machinery can greatly improve work efficiency and reduce the workload of farmers,but the complex operating environment,low intelligence,and high cost have always limited its application and promotion.Aiming at the low intelligence of agricultural machinery and the low utilization rate of land,this paper studies the path planning of the full coverage of agricultural machinery;the path tracking algorithm is improved for the problem of complex operating environment and difficult to precise control.The work of this paper includes the following aspects:1.Research on the full coverage path planning method of agricultural machinery,aiming at the problem of low land utilization rate,proposed a full coverage path planning method with the narrowest turning boundary.Firstly,the width model of the turning area is constructed based on the angle between the work path and the boundary of the plot and the minimum turning radius of the agricultural machinery.Taking the minimum sum of the width of each boundary turning area as the optimization goal,using the angle between the working path and the farmland boundary as the inequality constraint,using genetic algorithm to find the best working direction and the narrowest turning area.2.For the planning of the turning path,the shortest turning path method based on the fourth-order Bezier curve is proposed to solve the problem of long turning paths and sudden changes in curvature of the traditional turning methods.First,the shortcomings of the traditional turning path are analyzed,and then the fourth-order Bezier curve is used to model the turning path.Considering the problem of sudden curvature,the range of control points is limited.The fitness function is established with the constraints of the maximum curvature and the shortest turning path,and the path is searched by the elite genetic algorithm.Finally,the shortest turning path that satisfies the constraints of agricultural machinery dynamics is obtained.3.Aiming at the complex working environment of agricultural machinery,the traditional Pure Pursuit algorithm preview point selection is not robust,and can not be adjusted adaptively,a pure pursuit path tracking algorithm improved by reinforcement learning is proposed.Proximal Policy Optimization(PPO)is selected as the Deep Reinforcement Learning(DRL)algorithm,and combined with the Pure Pursuit(PP)method to construct the vehicle control architecture.The pure tracking method is used to generate the reference steering control command,and the PPO is used to derive the correction command to reduce the error caused by the improper selection of the preview point in the pure tracking algorithm.The fusion of these two controllers makes the overall operation more rob ust and adaptive,and obtains the best choice for improving tracking performance.Experiments show that the full coverage planning method in this paper improves the land use rate,and the turning path planning reduces the vehicle turning cost.The control algorithm improves the robustness and adaptive ability of the traditional tracking algorithm.