LiDAR-Inertial based Localization and Perception for Indoor Pursuit-Evasion Differential Games

Pursuit-evasion (PE) differential games are widely studied due to their enormous application prospects in robotics. Most of current works, however, focus on the theory and simulation, while even for some experiments the global information is directly acquired by external devices or just by sharing, which is far from the real scenarios. Motivated by this limitation, we propose a LiDAR-inertial based localization and perception method to estimate robot states such as position, pose and velocity required by high-level strategies. The LiDAR-inertial odometry (LIO) that utilizes data from 3D LiDAR and IMU provides self localization in an unknown environment. To locate other robots, a curve fitting based algorithm for 3D LiDAR point clouds is proposed, after which a sliding window average method is adopted to filter the noise in the perception results. Except for 3D LiDAR and IMU, no other sensors or communication is needed. Experiments are presented to illustrate the results.