Our research focuses on optimizing the human machine interface with robotic prostheses and exoskeletons. Much of this involves learning control methods to allow a human user to efficiently operate and work in sync with an assistive device. This involves extracting user intent through sensor fusion and machine learning algorithms along with other states such as gait phase, locomotion mode and walking speed. We then optimize robotic walking profiles to assist the human user. We focus our application with human user tests on individuals with walking disability such as amputees, stroke subjects, elderly subjects and others.