Role control method, model training method and related devices

Abstract

The invention discloses a role control method, which comprises the steps of obtaining first state information of a target role at a first moment, the target role comprising M joints, and the first state information being used for representing information of the M joints; based on the first state information, first action information of the target role at a second moment is obtained through a stateprediction model, and the first action information is used for representing information of (M-1) joints; a first moment is obtained according to the first action information, and the first moment comprises the moment of each joint in (M-1) joints at the second moment; and controlling the target role to execute the first action through the physical engine based on the first moment. The invention further discloses a model training method. The moment acting on each joint of the role can be calculated according to the action information, and then the role is controlled based on the moment in thephysical engine, so that the vivid action sequence is generated, and the action effect of the role is improved.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to a character control method, a model training method and related devices. Background technique [0002] With the continuous development of artificial intelligence in recent years, reinforcement learning has received more and more attention in the field of action imitation. Applying reinforcement learning technology to the action generation of game characters will help to optimize the efficiency of game development and improve the action effects of game characters. [0003] Nowadays, after designing the characters in the animation, an animator needs to give these characters different actions, such as running, walking, jumping and attacking. Animators need to pre-design a large number of animation clips through animation production software. During the game, an animation clip of the current scene can be played. [0004] However, the characters in the animation cli...

AI Technical Summary

Problems solved by technology

[0004] However, the characters in the animation clips often lack the simulation of the physical laws in the real world, and some actions may not conform to the real situation. For example, order the character to walk towards the wall. When the character is close to the wall, the character will stick to the wall.

As a result, it is difficult for the character's movements to achieve a real and natural state, which reduces the effect of the character's movements

Images