Motion determination model training method and device for virtual object, equipment and medium

Abstract

The invention provides a motion determination model training method and device for a virtual object, equipment and a medium, and belongs to the technical field of artificial intelligence. The method comprises the steps of determining a computing environment state of a virtual scene after a target duration based on a first environment state of the virtual scene; determining internal reward information according to the calculation environment state and the actual environment state of the first environment state at the next moment; adjusting parameters of the current action determination model according to the internal reward information; and in response to the fact that the current action determination model meets the first target condition, determining the current action determination modelas a trained motion determination model. According to the scheme, the motion output by the motion determination model can correspond to different game strategies, and the confrontation capacity and robustness of the virtual object on the game strategies are improved.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, in particular to a method, device, equipment and medium for training a virtual object's action determination model. Background technique [0002] With the development of artificial intelligence (AI) technology, AI has challenged top humans in various fields, and has approached the top level of human competition. For example, in the field of Go, Alpha Go defeated the world champion of Go, and in the field of games, Alpha Star defeated professional players of StarCraft II (a real-time strategy game) and so on. Currently, research on game AI issues has become a testing ground for exploring real-world general artificial intelligence. [0003] At present, for Multiplayer Online Battle Arena (MOBA), because MOBA games are affected by complex content such as lineup combinations, strategic goals, and tactical execution, reinforcement learning (Reinforcement Learning, RL) is usu...

AI Technical Summary

Problems solved by technology

[0004] The technical problem in the above technical solution is that since the reward signal of the AI ​​in the above RL is a dense reward signal defined by the technician, the trained AI can only execute a single game strategy, resulting in the AI’s ability to fight against the game strategy. Weak, lack of robustness

Images