Algorithm for Controlling an Autonomous Vehicle for Agriculture

This article presents a highly autonomous vehicle (HAV) control algorithm for agriculture. Using HAV in agriculture can reduce costs and the impact of chemicals on machine operators and increase labor productivity. Article describes the approach and technical aids for converting existing vehicles into HAV. Article discusses the considered proposed architecture of the HAV control subsystem. An example used in the article is the UAZ Patriot car. Proposed approach is portable to various types of vehicles. The article also presents an algorithm for generating the optimal trajectory of the HAV movement across the field, considering its boundaries and static obstacles. Proposed algorithm is based on reinforcement learning. The resulting trajectory gets loaded into the HAV control subsystem. Algorithm for modifying the trajectory of the HAV movement based on Reeds-Shepp curves is used to avoid dynamic obstacles. This algorithm allows to change the trajectory of the HAV to avoid dynamic obstacles and return to the original route. The YOLOv8 model processes the data from technical vision tools for object detection. Authors developed a simulator based on the Robot Operating System library to reduce costs and increase the speed of developing and debugging the HAV control algorithms. The framework of the research work considered the involvement of students from the Faculty of Mechanical Engineering and the Faculty of Information Systems and Technologies at the Ulyanovsk State Technical University.