Development and Evaluation of a Data Glove-Based System for Assisting Puzzle Solving.

Area of Science:

• Robotics
• Human-Computer Interaction
• Wearable Technology

Background:

• Automating complex, hands-on tasks is challenging due to dexterity requirements.
• Existing data gloves primarily focus on gesture recognition, not step-by-step task guidance.
• A need exists for systems that provide real-time, tactile feedback for task support.

Purpose of the Study:

• To develop and evaluate a tactile-sensing operation support system for multi-step manipulation tasks.
• To assess the system's effectiveness in guiding users and verifying task steps.
• To quantify improvements in task performance and user experience.

Main Methods:

• Developed an e-textile data glove with 88 pressure sensors and a tactile pressure sheet.
• Implemented a Convolutional Neural Network (CNN) for classifying grasped object states.
• Designed a Graphical User Interface (GUI) for step-by-step instructions and feedback.
• Conducted a user study using the Tower of Hanoi task as a proxy for manipulation.

Main Results:

• The CNN achieved 93.3% accuracy in classifying grasped states.
• The system reduced mean solving time by 51.5% and disc movements by ~20.
• Perceived workload (NASA-TLX) decreased by 53.1%, with a System Usability Scale (SUS) score of 75.

Conclusions:

• Tactile-based step verification and guidance are feasible for controlled multi-step tasks.
• The developed system shows potential for enhancing human performance in dexterous manipulation.
• Further research with diverse participants and tasks is needed for broader generalization.