Research On The Operation Method Of Deck Support Personnel Based On Reinforcement Learning

The issue of the carrier aircraft support operation is one of the key issues in carrier research,and the efficiency of support operations is also an important factor limiting the rate of carrier aircraft sortie recovery.However,there are few studies on support personnel operations.As for the process of carrier aircraft support,the support operation is completed by the support personnel,and all many kinds of factors that are ignored in the carrier aircraft support process are also handled by the support personnel.Therefore,it is necessary to study the operation method of the support personnel,so that the overall support process can be carried out in an orderly manner and the efficiency of the support personnel can be improved,which is necessary to enhance the carrier’s combat capability.In consideration of the emergency,the overall shipboard aircraft support process is sorted out and analyzed.The overall process is established as a Markov decision process model by capturing the operational characteristics of the support personnel and matching the “one-pit” support model and the integrated joint human-aircraft support model.To design a multi-agent reinforcement learning-based algorithm for operation personnel scheduling with emergency.The multi-agent framework is designed according to the heterogeneous characteristics of the support personnel,and the prioritized experience replay mechanism for emergency and adaptive conflict penalty coefficient mechanism are designed to optimize the training process and the results.Through simulation tests,it is proved that the designed algorithm can complete the support personnel scheduling while maintaining a good response to emergency.After the completion of the scheduling,the support personnel need to be involved in the job scheduling problem when they perform their operations.A job scheduling optimization algorithm based on Soft Actor Critic is designed for personnel fatigue.The reward shaping is applied to avoid the influence of sparse reward environment.To enable it to adapt to different workloads quickly,a workload fast adaptation mechanism is designed.It is verified by simulation tests that it can optimize the job scheduling to some extent.The advantages and disadvantages of the proposed personnel scheduling algorithm as well as the job scheduling optimization algorithm are further analyzed by comparing them with the traditional intelligent algorithm and analyzing the characteristics of their solutions.And through their own comparisons,it is shown that the prioritized experience replay mechanism for emergency and the adaptive conflict penalty coefficient mechanism have a positive impact on the training process respectively.And both of them jointly maintain the algorithm to obtain the final optimization results.The effectiveness of the workload fast adaptation mechanism is verified under different workloads,and its characteristics as well as advantages and disadvantages are analyzed to clarify its scope of application.