Decentralized multi-agent POMDPs framework for humans-robots teamwork coordination in search and rescue

Despite the fact that robots have reached a high level of autonomy in recent years, the need for human presence in certain situations is still essential, especially in search and rescue operations. The human extends the robots capabilities beyond of what they are capable of with current technologies. While current robotic devices are able to navigate, locate, and map search and rescue areas, some interventions require high degree of dexterity and information exchange that implies cooperation between the agents intervening in the area - human and/or robots. This paper presents a framework for modelling the coordination between human responders and robots in search and rescue scenarios using Decentralised Multi-agent Partially Observable Markov Decision Processes (Dec-MPOMDP). In this framework the human is treated as an intelligent agent with separate observations and actions that are communicated with the remaining team (human and robots) to reach the level of synergy required to accomplish joined tasks.