DynaSLAM II: Tightly-Coupled Multi-Object Tracking and SLAM

The assumption of scene rigidity is common in visual SLAM algorithms. However, it limits their applicability populated real-world environments. Furthermore, most scenarios including autonomous driving, multi-robot collaboration and augmented/virtual reality, require explicit motion information the surroundings to help with decision making understanding. We present this paper DynaSLAM II, a system for stereo RGB-D camera configurations that tightly integrates multi-object tracking capability. II makes use instance semantic segmentation ORB features track dynamic objects. structures static objects are optimized jointly trajectories both moving agents within novel bundle adjustment proposal. 3D bounding boxes also estimated loosely fixed temporal window. demonstrate does not only provide rich clues understanding but can be beneficial tracking.