Training method and training network for deep decision-making of UAV in strong confrontation environment

Abstract

The invention provides a training method and a training network for deep decision-making of UAVs in a strong confrontation environment, belonging to the technical field of deep decision-making of UAVs. The training network includes an input layer, a hidden layer, an output layer, a reward value acquisition module, a memory bank and a gradient training module. The training method or training network breaks through the limitation that the traditional UAV training method cannot adapt to the environment to generate a plan, so that the trained UAV can flexibly make autonomous decisions in a complex and changeable strong confrontation environment.

Description

technical field [0001] The invention relates to the technical field of UAV depth decision-making, in particular to a training method and training network for UAV depth decision-making in a strong confrontation environment. Background technique [0002] Most of the existing UAV decision-making methods do not consider the strong confrontation environment of UAVs, and cannot solve the autonomous decision-making problem of UAVs in a strong confrontation environment. For example, the paper "A UAV Autonomous Robust Decision-Making Method Based on Scenario Construction" uses an uncertainty solution method based on influence diagrams, which has the following two deficiencies: First, the decision-making method is essentially based on candidate solutions In the process of selecting the scheme with the greatest utility, the dimensions of emergencies that can be covered by the candidate scheme directly determine the robustness of the method. However, these candidate schemes can only be ...

AI Technical Summary

Problems solved by technology

[0002] Most of the existing UAV decision-making methods do not consider the strong confrontation environment of UAVs, and cannot solve the autonomous decision-making problem of UAVs in a strong confrontation environment.

For example, the paper "A UAV Autonomous Robust Decision-Making Method Based on Scenario Construction" uses an uncertainty solution method based on influence diagrams, which has the following two deficiencies: First, the decision-making method is essentially based on candidate solutions In the process of selecting the scheme with the greatest utility, the dimensions of emergencies that can be covered by the candidate scheme directly determine the robustness of the method. However, these candidate schemes can only be summarized from historical battle cases, and cannot be used for unexperienced situations. Decision-making under the scene; second, the influence diagram method needs to build the influence diagram model in advance, and it cannot be continuously improved after the model is built, and it is difficult to adapt to complex and changeable highly dynamic battlefield conditions

Images