Animation simulation method of AR augmented reality large screen interaction based on deep reinforcement learning

Abstract

The invention discloses an AR augmented reality large-screen interactive animation simulation method based on depth reinforcement learning, which collects action data of professionals and divides thedata as a reference action set. Firstly, two caffe convolution neural network frameworks are constructed, in which the states, actions and targets of animation characters are used as the first input,and the states, actions, targets and actions of lower limbs with wide range of motion are used as the inputs of the second grid, which is used to assist each other with the first network to acceleratethe learning rate. Animated characters are driven by PD controller, embedded in AR augmented reality system directly, or combined with the original animated characters in AR augmented reality system.The invention constructs a virtual animation character, through the reward and punishment information fed back to the character, makes the character know whether its behavior is correct or not, and through long-time study, the final character will have the ability of autonomous decision-making.

Description

technical field [0001] The invention relates to an animation simulation method, in particular to an animation simulation method for AR augmented reality large-screen interaction. Background technique [0002] As a publicity tool, AR large-screen interaction has an irreplaceable advantage in places with large traffic such as exhibitions, press conferences, and shopping malls. At the technical level, AR large-screen interaction is involved in various technologies such as image recognition, face recognition, state recognition, and gesture recognition. In terms of interactive methods, a series of interactive forms such as extended AR interactive games, AR interactive photography, etc. have performed well. [0003] At present, the methods for making motions of computer animation characters are mainly to manually draw key frames and use motion capture data for animation characters. The method of manually drawing key frames requires frame-by-frame design, which is time-consuming ...

AI Technical Summary

Problems solved by technology

But because of these design decisions, controllers are usually specific to a certain agent or task, and most of them are single-action training, and multi-action coherent training has not been applied in practice.

Images