Pedestrian Detection for Autonomous Mobile Robots using 3D LIDAR

There is a severe labor shortage due to the aging population, and there is an increasing demand for autonomous mobile robots such as delivery robots. These robots operate in the same space as pedestrians, so encounters and overtaking with pedestrians occur. In such cases, the robot can achieve safe and efficient path planning for both the robot itself and pedestrians by recognizing pedestrians as distinct from other obstacles. Therefore, this study aims to propose a pedestrian detection model and evaluate its performance. The sensor used in this paper is a 3D LIDAR, and pedestrian classification is conducted using a machine-learning technique: “Random Forest”. The“KITTI” dataset was utilized for model training and simulation. The evaluation of the proposed model showed an accuracy of approximately 98% in cluster-level tests.