Dexterous Manipulation Based on Prior Dexterous Grasp Pose Knowledge

Dexterous manipulation has received considerable attention in recent research. Predominantly, existing studies have concentrated on reinforcement learning methods to address the substantial degrees of freedom in hand movements. Nonetheless, these methods typically suffer from low efficiency and accuracy. In this work, we introduce a novel reinforcement learning approach that leverages prior dexterous grasp pose knowledge to enhance both efficiency and accuracy. Unlike previous work, they always make the robotic hand go with a fixed dexterous grasp pose, We decouple the manipulation process into two distinct phases: initially, we generate a dexterous grasp pose targeting the functional part of the object; after that, we employ reinforcement learning to comprehensively explore the environment. Our findings suggest that the majority of learning time is expended in identifying the appropriate initial position and selecting the optimal manipulation viewpoint. Experimental results demonstrate significant improvements in learning efficiency and success rates across four distinct tasks.

Authors

• Hengxu Yan
• Haoshu Fang
• Cewu Lu

Keywords

• Hands
• Adaptation models
• Accuracy
• Portable computers
• Refining
• Reinforcement learning
• Turning
• Force control
• Intelligent robots
• Dexterous Manipulation
• Grasp Pose
• Degrees Of Freedom
• Use Of Processes
• Hand Movements
• Efficient Learning
• Object Parts
• Robotic Hand
• Collision
• Time Step
• Task Completion
• Workspace
• Point Cloud
• Simulation Environment
• Joint Angles
• Robotic Arm
• Reward Function
• Real-world Experiments
• Real-world Environments
• Markov Decision Process
• Proximal Policy Optimization
• Imitation Learning
• Observation Space
• Goal Position
• Collision Detection
• Segmentation Dataset
• Increased Success Rate
• Knowledge Of Position
• High Degree Of Freedom
• Average Success Rate