Multistep Intent Estimation Guided Adaptive Passive Control for Safety-Aware Physical Human-Robot Collaboration.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Physical human-robot collaboration (pHRC) demands high safety and efficiency.
• Accurate human intent estimation and stable adaptive control are critical challenges in pHRC.

Purpose of the Study:

• To develop a novel two-loop adaptive passive control framework for safety-aware efficient pHRC.
• To improve human-robot disagreement reduction and robot assistance levels.

Main Methods:

• A two-loop framework with outer loop human intent estimation and inner loop adaptive passive control.
• Outer loop: Transformer-based Human Intent Estimator (THIE) with Conditional Variational Autoencoder (CVAE) for multistep predictions.
• Inner loop: Goal-Oriented Reinforcement Learning (GoRL)-based adaptive impedance control and Energy Tank-based Passive Model Predictive Control (ET-PMPC).

Main Results:

• The proposed framework significantly improves intent estimation accuracy and robot assistance levels.
• Experimental results demonstrate superior performance compared to state-of-the-art methods.
• The framework effectively balances disagreement and assistance while ensuring safety.

Conclusions:

• The novel framework advances safety-aware efficient physical human-robot collaboration.
• It offers a promising approach for real-time robot behavior adjustment based on human intent.
• The system enhances overall performance in human-robot interaction scenarios.