Multi-UAV task allocation based on GCN-inspired binary stochastic L-BFGS

Abstract

Task allocation has been one of the key issues for cooperative control of multiple unmanned aerial vehicles (Multi-UAVs), which has attracted a large number of researchers to conduct research in recent years. As the number of tasks and resource types increase, the solution time of most of the existing methods increases sharply, and are difficult to be deployed in other scenarios. To deal with task allocation problems with large-scale tasks and multiple types of resources, this paper proposed a multi-UAV task allocation method based on graph convolutional network (GCN)-inspired binary stochastic L-BFGS (GBSL-BFGS) with strong generalization. First, the objectives and constraints of the task allocation problem are analyzed, while a flexible and easily scalable method for describing the task allocation problem is proposed. Then, the GBSL-BFGS task allocation method is proposed for large-scale multi-UAV cluster. By introducing GCN as a graph mapper, the L-BFGS algorithm is able to optimize the binary decision matrix in the task allocation problem. Simulation experiments demonstrated that the GBSL-BFGS optimization method has a better performance and computational efficiency compared with other methods, especially for large-scale multi-UAV task allocation problems.