Game instruction model training method and device, storage medium and computing device

By moving the game instruction model to the front-end interaction module in the game environment and using a distributed architecture for training, the problems of latency and unstable training results caused by network latency are solved, and more efficient and stable virtual player action instruction generation is achieved.

CN116688506BActive Publication Date: 2026-06-19HANGZHOU NETEASE ZHIQI TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HANGZHOU NETEASE ZHIQI TECH CO LTD
Filing Date
2023-05-29
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

In the gaming environment, network latency in the backend computation process for calculating virtual player action commands causes delays and stutters, affecting the gaming experience of real players. Furthermore, traditional centralized model training frameworks cause training results to be affected by fluctuations in network conditions.

Method used

The game command model is moved to the front-end interaction module for local computation, and a distributed back-end training module is used for distributed training. Multiple service clusters are used to improve training efficiency, avoid network transmission latency, and enhance the stability of model training.

Benefits of technology

By using local front-end computing and distributed training, network latency issues are effectively avoided, improving the training efficiency and stability of the game command model, and enhancing the response speed of virtual player action commands and the model training effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116688506B_ABST
    Figure CN116688506B_ABST
Patent Text Reader

Abstract

This disclosure provides a method, apparatus, storage medium, and computing device for training a game command model. The method includes: receiving game data, wherein the game data includes virtual player state data; inputting the state data into a game command model mounted on a front-end interaction module to obtain action commands output by the game command model; and inputting the state data and action commands as training samples into a back-end training module, so that the back-end training module trains the game command model based on the training samples and iteratively generates an updated game command model.
Need to check novelty before this filing date? Find Prior Art