Learning-based and easy-to-generalize large-scale multi-agent collaborative game method

• Main Authors: HUANG BOMIN, LIN SHIWEI, ZHANG SHUANG
• Format: Patent
• Language: Chinese; English
• Published: 30.05.2025
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides a learning-based and easy-to-generalize large-scale multi-agent collaborative game method. The method comprises the following steps: S1, constructing a training sample and a newly added sample; s2, designing a prediction network for each defense agent based on a multi-layer perceptron; the network input is a newly added sample, and the output is a predicted game result; s3, designing a strategy network for each defense agent based on a graph neural network and a multi-layer perceptron; the network input is a prediction network output result, the position and speed information, sensed by each defense agent, of a part of attacker agent and the position and speed information of other parts of defense agents, and the network output is probability distribution of defense tasks of each defense agent; s4, on the basis of probability distribution, according to greedy sampling, a defense object for defending the subject is confirmed; and S5, realizing pursuit capture of the defense subject on th