Modeling method and action prediction system for multi-agent synchronous game

Abstract

The present invention provides a multi-agent synchronous game-oriented modeling method and an action prediction system. The modeling method includes: obtaining the state set and action set of the synchronous game problem; designing game features and actions according to the characteristics of the synchronous game problem. Feature encoding method; according to the game feature and feature encoding method, perform data preprocessing on the state set and action set to obtain the basic feature map and action decision map; establish a multi-scale feature fusion synchronous game strategy model based on a deep neural network ; According to the basic feature map and the action decision-making map, the synchronous game strategy model is trained layer by layer to obtain a well-trained synchronous game strategy model, while improving the game ability of the synchronous game strategy model, it is more powerful. real-time. The action prediction system realized based on the synchronous game strategy model has high accuracy and strong real-time performance, and has good practical application value.

Description

technical field [0001] The invention relates to the field of machine games, and more specifically, relates to a modeling method and an action prediction system for multi-agent synchronous games. Background technique [0002] Machine game research on how to allow computers to simulate humans to play games is one of the most challenging research directions in the field of artificial intelligence. Among them, the simultaneous game problem (where participants act simultaneously) represented by real-time strategy games has become a research hotspot in the field of artificial intelligence. [0003] The task of multi-agent synchronous game strategy is to let the computer accurately give the best actions of all participating units. The current mainstream solutions are search methods and multi-agent reinforcement learning. The search method builds a search tree through state clustering and action clustering, and calculates the local optimal solution online; multi-agent reinforcemen...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present invention provides a multi-agent synchronous game-oriented modeling method and action prediction system to solve the problem that existing modeling schemes cannot improve the game ability of the strategy model and have good real-time performance question

Images