Adaptive robot guidance through real-time compliance estimation and dual-modal control

Area of Science:

• Robotics
• Human-Computer Interaction
• Motor Learning

Background:

• Human instructors naturally blend physical touch and verbal cues for teaching motor skills.
• Robots providing only physical guidance can be perceived as intrusive, lacking contextual communication.

Purpose of the Study:

• To develop a robot guidance controller that mimics human instructors by coordinating physical and verbal guidance.
• To adaptively balance physical force and verbal instructions based on real-time human compliance.

Main Methods:

• An estimator infers human physical and verbal compliance from tracking errors.
• An optimization method dynamically allocates guidance between force and language modalities.
• A force-to-language model generates contextually relevant verbal feedback.

Main Results:

• Adaptive guidance significantly outperformed single-modality and fixed-combination approaches.
• Achieved up to 50% reduction in tracking error, 39% improvement in movement smoothness, and 27% faster task completion.
• User studies (N=12) validated the effectiveness of the adaptive coordination.

Conclusions:

• Coordinated physical and verbal guidance enhances human-robot collaborative learning.
• The adaptive system improves learning efficiency and task performance in motor skill acquisition.
• The approach is applicable to rehabilitation and general human-robot collaborative learning scenarios.