Related Experiment Video
Updated: May 16, 2026
10:51
An Experimental Platform to Study the Closed-loop Performance of Brain-machine Interfaces
Published on: March 10, 2011
Stochastic optimal control as a theory of brain-machine interface operation.
Manuel Lagang1, Lakshminarayan Srinivasan
1Neural Signal Processing Laboratory, Department of Radiology, University of California Los Angeles, Los Angeles, CA 90095-7437, USA. laganojunior@gmail.com
Neural Computation
|November 15, 2012
Summary
We developed a new model for brain-machine interfaces (BMI) that explains how the brain controls movements using optimal control theory. This model naturally predicts key phenomena observed in BMI experiments, offering insights into neural control mechanisms.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Control Theory
Background:
- Closed-loop brain-machine interfaces (BMI) are crucial for studying neural control and have clinical potential for neurological disorders.
- Existing BMI systems lack a fundamental, experimentally validated theoretical framework for their operation.
Purpose of the Study:
- To propose a compact theoretical model for closed-loop BMI operation based on stochastic optimal control.
- To describe how the brain controls BMI movements with sensory feedback and neural noise.
Main Methods:
- Developed a compact model using stochastic optimal control theory.
- Simulated goal-directed BMI movements with sensory feedback and noisy neural signals.
- Analyzed model predictions against experimentally validated phenomena.
Main Results:
- The model naturally reproduces phenomena like performance decline with bin width and compensation for decoder bias.
- Identified potential neural mechanisms for observed behaviors, such as energetic costs for bias compensation.
- Predicted that tuning curve shifts reflect the brain adopting new control policies during BMI skill mastery.
Conclusions:
- The proposed model offers a theoretical foundation for understanding brain-machine interface control.
- Provides testable predictions and insights into neural mechanisms underlying BMI performance.
- Suggests a framework for designing future BMI algorithms that leverage human-in-the-loop control.
