Reinforcement learning method for multi-fingered hand operation based on human motion prediction model

By employing a reinforcement learning method for multi-finger dexterity hand operation based on a human motion prediction model, the trajectories of human hands and objects are predicted simultaneously, a general reward function is constructed, and randomization techniques and curriculum learning are introduced into the simulation environment. This solves the problems of low training efficiency and poor generalization in multi-finger dexterity hand operation, and achieves efficient and stable anthropomorphic operation.

CN121492070BActive Publication Date: 2026-04-17ZHEJIANG UNIV
2 Cites 0 Cited by

Patent Information

Application Number
CN202610037406.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-01-13
Publication Date
2026-04-17
Estimated Expiration
2046-01-13

AI Technical Summary

Technical Problem

Existing multi-finger dexterity hand manipulation methods suffer from slow training convergence speed, unstable success rate, poor generalization to different types of objects, different poses and tasks, and neglect the perception and understanding of the manipulated object.

Method used

We employ a multi-finger dexterity hand manipulation reinforcement learning method based on a human motion prediction model. By training the human motion prediction model to simultaneously predict the future trajectories of human hands and objects, we construct a general reward function and introduce randomization techniques and curriculum learning in a simulation environment to improve training efficiency and success rate.

Benefits of technology

It significantly improves the versatility, convergence speed, and success rate of multi-finger dexterity hand operation, simplifies the application process, enhances the naturalness and stability of human-like operation, and enables efficient deployment in real-world environments.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application discloses a multi-fingered hand operation reinforcement learning method based on a human motion prediction model and belongs to the field of humanoid robot dexterous hand body operation. A human motion prediction model is trained by using data with human hand and object three-dimensional pose annotation, which is used to generate three-dimensional trajectories of human hand and object interaction in a generalizable manner. A general reward function is constructed by using the trajectory, which is used to train a dexterous hand reinforcement learning strategy in simulation. Then, the migration ability of the strategy from simulation to reality is improved through simulation parameter domain randomization and curriculum learning, and the trained reinforcement learning strategy is deployed on a real-world robot. The scheme introduces the understanding of the future pose of the object to be operated, predicts the interaction process of the human hand and the object as a whole, and constructs a unified, simple, efficient reward function based on this, which is suitable for different operation tasks and different types of dexterous hands, and has the advantages of strong generalization, high success rate and strong reward function universality.
Need to check novelty before this filing date? Find Prior Art