Robots that teach you how to walk

Is it possible to train people up to the point where you permanently overwrite the way they walk or swing a golf club? Professor Sunil K. Agrawal Ph.D., from the Mechanical Engineering Department of the University of Delaware recently gave a seminar talk in KAIST showing that it is indeed possible to change - to a certain extent - a deeply embedded behavior, such as the human gait, with some help from the assistive robotic technology.

Professor Agrawl's idea is conceptually very simple: take a group of patients who have suffered a stroke and have only partially recovered. Those patients usually walk more slowly than average, have an asymmetric gait, lack of ankle flection with consequent foot drop-toe dragging and are traditionally rehabilitated by human personal trainers. Repeatedly train them using a robotic exoskeleton which, by means of a haptic force feedback, constrains the trajectory of the walking feet within a virtual tunnel (see the picture below) with the shape of an input template gait. Eventually after some training, patients learn a new way of walking and this is ascertained by tracking and analyzing the data from how the patients walk after the training, without any aid from the robotic exoskeleton. The final experience for the users is very similar to that of a human trainer manually forcing the patients' legs to follow a specific trajectory, but the advantage of using machines is in the accuracy, consistency and continuity of the exercise.

What is more surprising though is that the same training works not only with patients who are recovering from a stroke, but also with healthy users of pretty much any age. In fact, Dr. Agrawal and his crew also tested their rehabilitation technique with healthy users for few weeks, training them daily with a template of a gait different from their own: the results showed that not only could they quickly learn a new way to walk, overriding their original gait, but that also the effects of the training lasted over time and that they could potentially be permanent if the treatment was continued (though further research is required to assess this).

Naturally, the next questions we would ask would be regarding other possible applications of this rehabilitation technique: can this technique be extended to more complex motion-based tasks, such as swinging a golf club or tennis racket, or learning how to play the piano? The possibilities are enormous and the exploration has just begun.

- Further readings:
Dr. Agrawal Mechanical System Laboratory and his research on rehabilitation robotics.