Aerial cooperative multi-robot system for remote sensing and automated victim search

This paper presents a distributed coverage control scheme for initially rapid scanning of the environment and then automated victim search in a large-scale firefighting scenario using a cooperative multi-agent system. The quick and efficient reaction of the disaster recovery team can significantly decrease damages and losses. Therefore, a comprehensive map of the affected area immediately after disaster occurred, leads to quickly initial assessment of damages, economic losses and casualties, which can highly improve the efficiency of further relief actions. We considered a firefighting scenario in which fixed-wing UAVs initially fly over the affected area and accomplish a generally and quick scan to identify the critical sections where the fire caused more damage and had a high intensity, severity and spread rate. Then, multi-rotors hover as close as possible to exactly locate the victims and survivors or also damaged infrastructure for more detailed information, and then, firefighters can plan the rescue mission more efficiently. We applied a distributed coverage control scheme so that a heterogeneous multi-agent system cooperatively could explore the affected area and search for the exact location of probable victims or survivors. To achieve this goal, UAVs collaboratively build the cognitive maps including target probability map (TPM) and uncertainty map (UM). Then, based on cognitive map and also their prediction of the next moves of themselves and their neighboring agents as well, UAVs make decision about their optimal collision-free paths. The efficiency of the implemented algorithm is demonstrated by comparison between cooperative and non-cooperative methods. The results indicate that in cooperative distributed scheme, coverage percentage and global average uncertainty converge to their admissible maximum and minimum value in a considerably less time, respectively.

Automated summary

This paper presents a distributed coverage control scheme for rapid scanning and victim search in a large-scale firefighting scenario using a cooperative multi-agent system. The efficiency of the scheme is demonstrated through comparison between cooperative and non-cooperative methods.