3D INTRA-GROUP LOCALIZATION AND MAPPING WITH MULTIPLE HETEROGENEOUS ROBOTS

3D Mapping with multiple heterogeneous robots that work as a team is a concerted effort to learn its surrounding environment in a quick and efficient way. Intra-group localization of multiple robots in the map generated by sensing the environment is a supplement to the mapping process. Each robot receives 3D information from multimodal sensors like Laser Range Finders (LRF), Perspective Camera, Inertial Navigation Sensors (INS) and odometers. Fusion of sensor data from these sensors into a world frame is a twofold problem. Initially, the multimodal sensor data is fused into respective robot coordinate frames. Secondly, sensory data of each robot is fused into a global coordinate frame, given the relative poses of multiple robots. The former involves intra-sensor calibration. The latter involves registration of two image inputs from the camera or laser scans in 3D (taken at reference and current positions of the mobile robot), with the help of odometry and orientation from INS. We review the cooperative localization techniques and treat laser registration as a separate problem in localization. We consider obtaining relative poses of multiple robots by fusing vision and laser scan data and discuss their efficiency and robustness. Efforts to generate virtual models of the robot?s surrounding environment in 3D that contain both depth and texture using laser scan maps and vision input is reviewed.