Game action processing method and device

Abstract

The embodiment of the invention provides a game action processing method and device. The method comprises the steps of acquiring the current game state and action space of a non-player character; inputting the current game state into a pre-trained reinforcement learning network model to obtain a reinforcement strategy, wherein the reinforcement strategy comprises a first selection probability of each game action; inputting the action space into a pre-trained auxiliary rule network model to obtain an auxiliary strategy, wherein the auxiliary strategy comprises a second selection probability ofeach game action; determining a target strategy according to the reinforcement strategy and the auxiliary strategy, wherein the target strategy comprises a target probability generated by each game action on the basis of the corresponding first selection probability and the corresponding second selection probability; screening out a target game action from the action space according to the targetprobability of each game action and controlling the non-player character to execute the target game action. The game action processing method and device can improve the game experience of players.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a game action processing method and a game action processing device. Background technique [0002] Reinforcement learning, also known as reinforcement learning and evaluation learning, is an important machine learning method that has many applications in the fields of intelligent control robots and analysis and prediction. The policy network (Policy Network) trained through reinforcement learning, the so-called policy network, is to establish a neural network model, which can directly predict the most current policy (policy) by observing the state of the environment. Execution of this policy can obtain the current and the maximum expected return value (reward) in the future. [0003] Reinforcement learning has a wide range of application scenarios in games. The policy network trained through reinforcement learning has higher intelligence in games than the strategies manu...

AI Technical Summary

Problems solved by technology

However, relying solely on reinforcement learning cannot learn the game rules in the game, which will lead to decision-making errors

Images