Air combat maneuvering strategy generation technology based on deep random game

Abstract

The invention discloses a short-distance air combat maneuvering strategy generation technology based on a deep random game. The technology comprises the following steps: firstly constructing a training environment for combat aircraft game confrontation according to a 1V1 short-distance air combat flow, and setting an enemy maneuvering strategy; secondly, by taking a random game as a standard, constructing agents of both sides of air combat confrontation, and determining a state space, an action space and a reward function of each agent; thirdly, constructing a neural network by using a maximumand minimum value DQN algorithm combining random game and deep reinforcement learning, and training our intelligent agent; and finally, according to the trained neural network, obtaining an optimal maneuvering strategy under an air combat situation through a linear programming method, and performing game confrontation with enemies. The thought of random game and deep reinforcement learning is combined, the maximum and minimum value DQN algorithm is provided to obtain the optimal air combat maneuver strategy, the optimal air combat maneuver strategy can be applied to an existing air combat maneuver guiding system, and effective decisions can be accurately made in real time to guide a fighter plane to occupy a favorable situation position.

Description

technical field [0001] The invention relates to the technical field of air combat game maneuver strategy generation, and particularly designs an air combat maneuver strategy generation technology based on deep random games. Background technique [0002] Air power is playing an increasingly important role in modern warfare, and the battle for air supremacy largely determines the outcome of a war. However, the air combat situation is changing rapidly, and the information that needs to be collected is extremely complex, making it difficult for the combatants to make decisions after sensing the air combat situation. Traditional methods cannot achieve a fast and accurate air combat strategy. The present invention proposes an air combat maneuver strategy generation technology based on deep stochastic game. The method combines the ideas of deep reinforcement learning and stochastic game, and can effectively select the maneuvering action that is beneficial to our side in real time a...

AI Technical Summary

Problems solved by technology

However, the air combat situation is changing rapidly, and the information that needs to be collected is extremely complex, making it difficult for the combatants to make decisions after sensing the air combat situation. Traditional methods cannot achieve a fast and accurate air combat strategy

Images