System and method for brain machine interface (bmi) control using reinforcement learning

A neural prosthetic system comprising:

• a micro-electrode array electro-chemically coupled to a neural structure of a subject to capture neural activity in the neural structure;
• a prosthetic device that performs one or more behaviors for the subject in accordance with the neural activity;
• and
• a Brain Machine Interface (BMI) agent operatively coupled to the micro-electrode array and the prosthetic device that monitors one or more states of the neural activity, receives feedback associated with a behavior of the prosthetic device, learns a functional mapping between the neural activity and the one or more behaviors in view of the feedback, and applies at least one control action to the prosthetic device in accordance with the learning to maneuver the prosthetic device for a targeted behavior.

A method for Brain Machine Interface (BMI) control, the method comprising:

• capturing neural signals in a neural structure of a subject during a mental task by the subject to control a prosthetic device;
• receiving feedback associated with a behavior of the prosthetic device responsive to the control;
• learning a functional mapping between the mental task and the behavior in view of the feedback;
• and
• applying at least one control action to the prosthetic device in accordance with the learning to control the prosthetic device for a targeted behavior.