We presented our paper “Situational Adaptive Motion Prediction for Firefighting Squads in Indoor Search and Rescue” at ICRA in the Long-Term Human Motion Prediction workshop.
Firefighting is a complex but poorly automated task. To minimize ergonomic and safety risks for firefighters, robots could be used in a collaborative approach. To enable human-robot teams in firefighting, important fundamentals are still missing. Among other things, the robot must predict human movement because occlusions are ubiquitous. In this work, we propose a novel motion prediction method for firefighting squads in indoor attack. Squad paths are generated using an optimal graph-based planning approach that represents firefighter tactics. Paths are generated on a per-room basis, allowing dynamic local adaptation of paths without global replanning. The movement of individual agents is simulated using a modification of the Headed Social Force Model. We evaluate the feasibility of the pipeline using a novel dataset generated from real footage and show its computational efficiency.
Contact: Elodie Hüsing
