"Body-In-The-Loop": Optimizing Device Parameters Using Measures of Instantaneous Energetic Cost.

Area of Science:

• Biomedical Engineering
• Robotics
• Human-Computer Interaction

Background:

• Assistive robotic devices (powered prostheses, orthoses, exoskeletons) require precise parameter tuning for optimal physiological performance.
• Real-time optimization with a human subject ('in-the-loop') is challenging due to noisy sensor data and dynamic delays.
• Existing methods for parameter optimization can be time-consuming and may not adapt well to dynamic physiological responses.

Purpose of the Study:

• To demonstrate and evaluate novel algorithms for the online optimization of assistive robotic devices.
• To estimate physiological objectives in real-time and identify optimal device parameters efficiently.
• To compare the performance of three distinct online optimization algorithms: Steady-State Cost Mapping, Instantaneous Cost Mapping, and Instantaneous Cost Gradient Search.

Main Methods:

• Developed algorithms combining dynamic estimation and response surface identification to handle noisy and delayed sensor data.
• Evaluated three algorithms (Steady-State Cost Mapping, Instantaneous Cost Mapping, Instantaneous Cost Gradient Search) with eight healthy human subjects.
• Used step frequency optimization to minimize metabolic energetic cost as a standardized, repeatable evaluation metric.

Main Results:

• All three tested algorithms achieved high accuracy in estimating optimal step frequency, with average errors less than 1% and standard deviations between 4-5%.
• Instantaneous Cost Mapping significantly reduced subject walking-time from over an hour to under 10 minutes.
• Instantaneous Cost Gradient Search demonstrated potential for efficient multi-dimensional parameter optimization, outperforming other methods in scalability.

Conclusions:

• The developed online optimization methods are accurate and efficient for tuning assistive robotic devices.
• Instantaneous Cost Mapping and Instantaneous Cost Gradient Search offer significant time savings and improved optimization capabilities, especially for complex systems.
• These algorithms provide a robust framework for real-time physiological objective optimization in human-robot interaction.