The MIT Biomimetics Robotics Laboratory’s HERMES humanoid robot demonstrates the validity and intuitiveness of human-inthe- loop control by having the robot mimic the human operator’s every movement. However, this exact mimicry is limited by the human’s speed and strength. The implementation of Hidden Markov Models, built using reinforcement learning techniques, will allow the robot to learn and recognize the human operator’s intended movements. Using this model, the robot will observe the beginnings of the operator’s motion, use the probablistic model to guess the intended action, and seamlessly transition from direct mapping into on-board optimized movements. This will allow dynamics calculations to occur under the hood, moving the human operator into more of a supervisory role

I work in the MIT Biomimetics Robotics Lab on controls and mechanical design. Last summer I worked in the Stanford Department of Bioengineering, implementing a computer vision based model- less control algorithm on a flexible robot for cardiac surgery. I also lead the mechanical division of the MIT Robotics Team, designing and building teleoperated and autonomous rovers for NASA competitions