Method and system for human-robot motion data mapping

Abstract

The invention provides a method for human-robot motion data mapping. The method includes obtaining training data which include human body motion data and robot sample data of corresponding motion; using the human body motion data as input, using the robot sample data as expected output, and obtaining a deep learning model by training a deep learning network, so as to obtain a mapping relation of human-robot motion data. The method provided by the invention realizes mapping of human-robot motion data based on a deep learning method, avoids plenty of repeated kinematics solving, and can drive a robot to perform motion more flexibly and accurately.

Description

technical field [0001] The invention relates to the technical field of robot learning, in particular to a method and system for human-robot motion data mapping based on deep learning. Background technique [0002] The precise motion generation of robots / humanoid robots has very important practical significance for the robot industry, animation production, medical health and other fields. For example, in wearable devices, through related algorithms, humanoid robots can simulate and generate motion data of users of different ages and different physical fitness levels, which can provide test data for later device debugging and algorithm evaluation, which facilitates the testing process It also saves the cost of testing. [0003] In the prior art, the technologies for motion generation of humanoid robots mainly include motion analysis method, motion capture data conversion method and machine learning method. [0004] The motion analysis method establishes a set of kinematic an...

AI Technical Summary

Problems solved by technology

However, as the scale of data increases and the complexity of robot construction increases, in practical applications, existing machine learning models are easily limited by the model itself. For different robotic devices, the adaptability is not high, and the generalization ability is not strong. , prone to bottlenecks such as underfitting and non-convergence

Images