Towards Collaborative Obstacle Avoidance using Small UAS in Indoor Environments

摘要

This paper proposes an intuitive and collaborative human-robot approach to perception and navigation for sUAS operating in unknown indoor environments, such as disaster response scenarios. The complexity of indoor environments coupled with the urgency of first responder missions make most available obstacle avoidance solutions too complex, unpredictable, impractical or ineffective. This paper presents a human-robot collaborative obstacle avoidance approach with real-time performance. Multiple RGB-D cameras are combined with a front-facing stereo pair to achieve robust localization and 3D perception around the aircraft. Obstacle avoidance is accomplished by dynamically prioritizing mapping data along the desired navigational path, projecting 3D obstacles onto a virtual plane, representing complex obstacle data with filtered geometric clusters and performing geometrical extrusion analysis to generate a collision-free solution. Collaboration with the Pilot-In-Command (PIC) is opportunistic and bidirectional allowing the PIC to control the aggressiveness of the obstacle avoidance solution on-the-fly easily adapting to the complexity of unknown indoor environments.