Design of obstacle detection method for autonomous driving in agricultural environments

Abstract

With the rapid development of the internet of things and information and communication technology, several studies into autonomous agricultural vehicles, such as self-driving tractors, drones, and seed-planting robots, have been undertaken. Autonomous farming systems have the potential to produce more crops with less impact on the environment and less effort, and self-driving agricultural vehicles are among the innovative technologies that could be key to future food supplies. In this study, we design an obstruction detection method based on point clouds, for autonomous driving in an agricultural environment. Pulsed LiDAR technology with a bandwidth of 1,550nm is adopted and the LiDAR sensor with an FoV of 90 degrees is utilized. We design a deep learning model to detect property information for structured or unstructured obstructions.