Obstacle detection system for autonomous vineyard robots based on passthrough filter

Abstract

This research aims to solve an obstacle detection problem to enable safety autonomous robot to work in complex vineyard environments. The problem remains challenging because paths planning to avoid obstacles discovered by onboard 3D LiDAR requires creating and updating a representation of the environment that can be searched by feasible paths. The process is computationally expensive. In this paper, we proposed a passthrough filter based obstacle detection solution in the robot operation system (ROS) architecture without increasing the hardware burden. In this solution, 3D LiDAR mounted on the robot was used to do the tree-row followed navigation and the obstacles detection in different ROS function packages with different point cloud processing. In the proposed solution, the range of interest (ROI) to detect the obstacle can be set by the user interface. The ROI is 0.7 m to 6m in front of the robot header. To verify the proposed solution, different types of obstacles including static small things like boxes, static big items like another robot and moving person were detected in field experiments. Experiments demonstrated that the proposed solution could detected obstacles in determined ROI successfully with low computational cost as 10 ms. Keywords: autonomous robot, 3D LiDAR, obstacle detection, obstacle Estimation DOI: 10.33440/j.ijpaa.20220501.192 Citation: Ran W X, Lan Y B, Dai X L, Gu J, Liu B Y, Geng L J, Han X and Yi L L. Obstacle detection system for autonomous vineyard robots based on passthrough filter. Int J Precis Agric Aviat, 2022; 5(1): 41–46.