Real-time 6DoF localization for a mobile robot using pre-computed 3D laser likelihood field

This paper proposes a real-time full-6DoF localization method designed for a mobile robot moving in human-inhabited environments. The field of robotics is expanding, and 6DoF pose estimation in the real world is an important function for robots. The proposed method is based on typical Monte-Carlo Localization algorithm, and we expanded it to 6DoF realtime pose estimation. For real-time calculation, the proposed method uses 3D Laser Likelihood Field generated from pre-computed 3D environment map. The 3D Laser Likelihood Field is represented voxels including the distance between its center and the nearest neighbor obstacle. To speed up the calculation of matching score between 3D environmental map and 3D-LIDAR point cloud, we apply it to the measurement process of Monte-Carlo Localization. The system is evaluated with MoCap and demonstrated in human-inhabited environments. As these experiments illustrated, the robot could estimate its self-pose robustly in real-time with the algorithm running on an ordinary notebook PC.