Objective Model Selection for Identifying the Human Feedforward Response in Manual Control.

Area of Science:

• Human-Computer Interaction
• Control Theory
• Systems Engineering

Background:

• Manual control tasks involve predictable targets and random disturbances.
• Human controllers (HC) likely use feedforward control for targets and feedback for disturbances.
• Identifying human feedforward control is challenging for existing system identification methods.

Purpose of the Study:

• To present an objective model selection procedure for identifying human feedforward responses.
• To develop a new model selection criterion for determining model order and the presence of feedforward control.
• To address the issue of false-positive feedforward detection in system identification.

Main Methods:

• Utilized linear time-invariant autoregressive with exogenous input (ARX) models.
• Proposed a modified Bayesian Information Criterion (BIC) with an added penalty for model complexity.
• Employed Monte Carlo computer simulations to test the method and determine appropriate weighting.

Main Results:

• The classical BIC incorrectly identified feedforward control in pure feedback systems (false positives).
• The modified BIC, with appropriate weighting derived from simulations, successfully eliminated false-positive feedforward detection.
• The proposed method accurately identified human controller dynamics across various control tasks.

Conclusions:

• The developed identification procedure effectively distinguishes between feedforward and feedback control in human manual tasks.
• The modified BIC offers a robust solution to the challenge of false-positive feedforward detection.
• Future work will involve validating the method with experimental human-in-the-loop data.