A Deadlock-Free Hybrid Estimation of Distribution Algorithm for Cooperative Multi-UAV Task Assignment With Temporally Coupled Constraints
Ruipeng Zhang, Yanxiang Feng, Yikang Yang, Xiaoling Li
Abstract
This article addresses the cooperative multiunmanned aerial vehicles task assignment problem (CMTAP) with temporally coupled constraints and aims to find a feasible assignment to minimize the equivalent distances of all tasks. We first present a mixed-integer linear programming model of CMTAP. To solve the undesirable deadlocks of CMTAP, a Petri net amender is constructed based on a candidate solution, and a deadlock-free solution is equivalent to a feasible transition sequence that can be fired sequentially in the corresponding amender. With this amender, we present a Petri net-based deadlock amending method (PDAM) with polynomial time complexity to convert a deadlocked solution into a deadlock-free solution. Also, a deadlock-free hybrid estimation of distribution algorithm (DHEDA) is developed for CMTAP by embedding PDAM into the original EDA. To further improve the solution quality, we establish a local exploitation method, and an adaptive operational probability is used to balance the computational burden and local exploitation ability. Then, a match-up-based reassignment method is proposed to cope with time-sensitive targets. Finally, extensive computational experiments demonstrate that PDAM is more effective at solving deadlocks than graph-based methods, particularly for large-scale CMTAP, and DHEDA outperforms existing algorithms when solving CMTAP.