Hierarchal Application of Receding Horizon Synthesis and Dynamic Allocation for UAVs Fighting Fires

This paper explores the design of a high-level mission planner and controller for managing unmanned aerial vehicles (UAVs) fighting a wildfire through the utilization of reactive synthesis and dynamic allocation of the UAVs as resources for the fire. The contribution of this paper is a study on the hierarchal integration of reactive synthesis, used for assuring desired system design traits, and dynamic allocation, used for making heuristic-based decisions. Reactive synthesis provides a formal means of guaranteeing the UAVs' transition to areas of fire, refill of water, and land as defined by the linear temporal logic specifications. Dynamic allocation coordinates the behavior of multiple UAVs through assignments to regions of fire based on a cost function that takes into consideration the fire locations relative to a UAV, distance to the domain edge, wind speed and direction, and the amount of suppressant already present. The use of receding horizons in the reactive synthesis formulation incorporates horizons defined only through spatial distance from a goal. Modifications to these horizon definitions guarantee that the scenario still maintains the overall realizability of the formal specifications after the inclusion of static obstacles. This paper shows the effectiveness of multiple UAV fleets in slowing down the progression of fires from reaching the domain edge through six fire scenarios. At last, our results and successful application demonstrate the utilization of reactive synthesis in larger task spaces and the implications of abstracting UAV transitions for use in formal methods.