Unmanned aerial vehicle combat analog simulation method based on multistage post experience playback

Abstract

The invention discloses an unmanned aerial vehicle combat analog simulation method based on multistage post experience playback. The method comprises an unmanned aerial vehicle simulation preparation information setting step, an unmanned aerial vehicle training network construction step based on multi-target agent training, an unmanned aerial vehicle training network training step based on multi-target agent training, and a repeated simulation and ending step. According to the method, multiple levels of post experience playback pools are set, three plot samples with different priorities are stored and utilized, random sampling is carried out according to different priorities of the experience playback pools and different probabilities, the higher the priority is, the larger the sampling probability is, useful sample information is utilized, the training efficiency is greatly improved, the overall utilization rate of the samples is improved, and the training efficiency is improved. And learning of the multi-target agent model is accelerated.

Description

technical field [0001] The present invention relates to the field of virtual simulation, in particular, to a combat simulation method for unmanned aerial vehicles based on multi-level post-event experience playback, which can introduce a multi-level post-event experience playback mechanism in the deep reinforcement learning of multi-target agents, and quickly improve the intelligence of intelligent agents. The learning performance makes it possible to quickly improve the speed of the agent learning to complete multi-objective tasks in the UAV flight simulation simulation. Background technique [0002] With the development of unmanned and intelligent technology, the use of drones has become an important topic in the fields of civil and military science. The initial drones were mainly operated manually. With the development of intelligence and simulation, A variety of intelligent agent simulation control methods have been applied to UAV flight simulation operations. [0003] ...

AI Technical Summary

Problems solved by technology

Although the deep reinforcement learning method based on post-event experience playback uses an automated learning method and improves the training efficiency of UAV navigation in a simulated simulation environment, this method uses random sampling to collect UAV flight samples for the intelligent agent. learning, resulting in low sample utilization and even affecting the final performance of the UAV agent

[0005] Therefore, how to solve the problem of insufficient utilization of UAVs in the process of multi-objective agent reinforcement learning training and improve the learning efficiency of UAV agents has become an urgent technical problem.

Images