Real-time vibrotactile pattern generation and identification using discrete event-driven feedback

Area of Science:

• Robotics and Human-Computer Interaction
• Biomedical Engineering
• Haptic Feedback Systems

Background:

• Robotic systems increasingly rely on sophisticated feedback mechanisms for intuitive control.
• Vibrotactile feedback offers a promising modality for conveying information in human-robot interaction.
• Real-time classification of sensor data is crucial for effective event-driven feedback.

Purpose of the Study:

• To assess human identification of vibrotactile patterns generated by real-time discrete event-driven feedback.
• To evaluate the performance of decision tree (DT) algorithms implemented on a field-programmable gate array (FPGA) for classifying robotic hand states.
• To determine the accuracy of human participants in recognizing sequential vibrotactile patterns.

Main Methods:

• Utilized force and bend sensor data from a robotic hand to predict movement and object states.
• Implemented DT algorithms on an FPGA for real-time classification of states.
• Six participants performed sequential pattern recognition tasks using vibrotactile feedback from two actuators (Haptuators).
• Vibrotactile stimuli were generated based on classified states, using varying frequencies and magnitudes mapped to specific events.

Main Results:

• The DT algorithm achieved 97% accuracy for object-type classification and 88% for movement-type classification in real time.
• Participants demonstrated low-to-medium accuracy in recognizing 14 discrete-event sequences.
• Performance for recognizing individual events within sequences was higher, with 78% accuracy, 76% recall, and 76% precision.

Conclusions:

• FPGA implementation of classification for discrete event-driven vibrotactile feedback is feasible for haptic devices.
• Vibrotactile feedback can effectively signal state transitions in robotic interactions.
• Additional contextual cues may further enhance the recognition of complex vibrotactile sequences.