Skip to Main Content
This paper presents a laser-based pedestrian tracking system employing a group of mobile robots. Each robot detects pedestrians in its own laser scan images using an occupancy-grid-based method. It then tracks the pedestrians using Kalman filter and global-nearest-neighbor (GNN)-based data association. It also generates a local map using EKF-SLAM. Using inter-robot communication, the robots located near each other exchange information on tracked pedestrians and local maps. Tracking data are fused using the covariance intersection (CI) method, and a global map is built by merging the local maps. In our tracking method, all robots share the information with each other; hence, they can always recognize pedestrians who are invisible to another robot. The method is validated by the results of both indoor and outdoor experiments in which two pedestrians are tracked by three mobile robots. Our tracking system functions in a decentralized manner without any central server.