Research on Motion Transfer Method from Human Arm to Bionic Robot Arm Based on PSO-RF Algorithm.

Area of Science:

• Robotics
• Biomechanics
• Machine Learning

Background:

• Existing bionic robot arm motion transfer methods struggle with dynamic compliance and real-time adaptability due to simplified models.
• Human-like motions in bionic arms require advanced techniques to overcome limitations of kinematic equivalence and dynamic models.

Purpose of the Study:

• To present a novel motion transfer method for bionic robot arms using a hybrid Particle Swarm Optimization-Random Forest (PSO-RF) algorithm.
• To improve joint space mapping accuracy and dynamic compliance in human-to-robot motion transfer.
• To enable real-time adaptability and complex human-like motions in bionic robotic systems.

Main Methods:

• Utilized a high-precision optical motion capture (Mocap) system for human arm trajectory recording.
• Applied Kalman filtering and Rauch-Tung-Striebel (RTS) smoothing to process motion data, followed by geometric vector analysis for initial joint angle computation.
• Developed a hybrid PSO-RF model trained on five action sequences to predict human joint angles, overcoming limitations of geometric analysis.

Main Results:

• The PSO-RF model achieved high prediction accuracy for joint angles, with R² = 0.932 for the elbow joint.
• The motion transfer system demonstrated real-time performance with a low latency of 0.1097 seconds.
• Experimental validation confirmed the effectiveness of the proposed method in complex human-like motions.

Conclusions:

• The hybrid PSO-RF method significantly enhances motion transfer accuracy and dynamic compliance for bionic robot arms.
• This framework addresses joint-level dynamic transfer challenges, promoting compliant human-robot interaction.
• The developed system offers a promising framework for applications in intelligent manufacturing and rehabilitation robotics.