People tracking by multiple ground LiDARs based on distributed interacting multimodel method

Abstract

This paper presents a cooperative people tracking using networked light detection and rangings (LiDARs) allocated in an environment. Each LiDAR detects people from the LiDAR-scan data using a background subtraction method and sends the people positions to the adjacent LiDARs. It estimates the people poses (positions and velocities), and the estimates are exchanged among the adjacent LiDARs. A distributed interacting-multimodel (DIMM)-based method is utilized to accurately estimate poses of people under various motion modes, such as stopping, walking, and suddenly rushing out, in a distributed manner without a central server. A global nearest neighbor (GNN)-based data asscosiation as well as a rule-based detection and track management is implemented to reduce false tracking in environments with people close to each other. The DIMM-based method works in any LiDAR network topologies, and therefore, this may provide a degree of scalability that cannot be achieved by conventional centralized interacting-multimodel (CIMM)-based method with a central server. Simulation results of people tracking by three LiDARs in an intersection environment reveal the tracking performance of the proposed DIMM-based method by comparison of conventional CIMM and Kalman filter-based methods.

Keywords

• people tracking
• lidar
• distributed state estimation
• distributed interacting multimodel estimator