LiDAR Based Road Detection and Control for Agricultural Vehicles

Abstract

Recently, the population of agricultural workers in Japan has been declining and aging. As a result, labor shortages have become a serious problem in the agricultural industry. However, the strong dependence on food imports has become a problem. To address this situation, it is necessary to increase the efficiency of food production by automating agricultural work. In this study, we focus on transportation in farming, and develop an unmanned transportation truck. Farm roads are often unpaved or otherwise uneven, and the surrounding environment changes dramatically depending on road surface conditions, vegetation, weather, and season. To realize an unmanned transportation system, achieving robust environmental recognition and operational control is important. Although it is almost certain that no single method can manage all situations for this goal, as part of the system, we propose a method for generating a target path for farming environments using 3D LiDAR (Light Detection and Ranging), and apply a control law to follow the path on uneven surfaces robustly. Furthermore, we generate two types of paths, one based on the center of the drivable area and the other based on road surface geometry, and integrate them into a target path based on their reliability. A nonlinear path-following control law is designed to follow the target path. The proposed method was applied to an experimental truck, and stability was achieved on a real farm road.