A new multi-domain cooperative resource scheduling method using proximal policy optimization

For the complex environment and massive multi-source data, the capability of multi-domain cooperative resource scheduling has become extremely important. Optimal scheduling can reduce operating costs and time, and MDLS is still the most commonly utilized algorithm in combat task scheduling today, despite of its defects. This research provides a plausible new method for the MDCRS problem, a resource scheduling method based on deep reinforcement learning (DRL), which has proven to be effective for other scheduling problems. Aiming at the resource scheduling problem in the multi-domain cooperative operation, under timing constraints, an MDCRS model is created using the shortest completion time as the objective function. On this premise, this paper presents an MDCRS-MDP model based on Markov decision processes, in which a two-dimensional action space that can simultaneously allocate action and match platform is designed and a dense reward function with strong connections to the criterion for sparse makespan minimization is provided. A resource scheduling approach utilizing DRL is proposed, including task-platform matching and task sequencing, based on the MDCRS-MDP model. Finally, combined with the joint landing operation, the experimental results verify the effectiveness of the proposed method for solving MDCRS and demonstrate the significant advantages over traditional dispatching rules and meta-heuristic optimization algorithms.