Task Allocation in Uncertain Environments using a QuadTree and Flow network

Systems of multiple UAVs have been used for surveillance and reconnaissance operations for the past few decades. One of the most challenging problems with deploying multiple UAVs with different capabilities is how to individually assign them to a set of tasks in such a way that optimizes the overall mission goal subject to a set of resource constraints. In this paper, we consider the problem of task collaboration and coordination between a team of UAVs using
a priori
reconnaissance about the task and risk distributions. We present an intuitive approach to allocating tasks with centralized and decentralized techniques using a Quad-Tree and K-Partite graph. The Quad-Tree formally models the reconnaissance using its quadrants as potential task locations. It is used to compute the utilities for assigning each UAV to the quadrants, taking into account the cost of reaching the quadrant centroid and eliminating the risk imposed by static enemies. The K-Partite graph performs task allocation and coordination between UAVs and the quadrants to maximize the collected intelligence, while maintaining the time, fuel, and safety constraints of the UAVs. Once the initial task assignments are centrally generated, each UAV is responsible for assigning itself to other tasks while exploring the field. The coordination of UAVs is achieved by communicating the changes of their local graph and tree. We assume that communication range and bandwidth are unlimited, but we describe how our approach can be extended for situations where communication is constrained. Simulation experiments demonstrate that our approach effectively distributes the UAVs over the environment and allows them to accomplish their mission within the resource limits.