In Caltech‘s Advanced Mechanical Bipedal Experimental Robotics Lab (AMBER), a team of graduate students is designing a new approach to generate gaits for robotic assistive devices.

The team is led by Professor Aaron Ames, Bren Professor of Mechanical and Civil Engineering and Control and Dynamical Systems. The research and development are aimed to guarantee better stability and more natural locomotion for different users.

A musculoskeletal model is a computational tool to estimate the connection between joint contact and muscle force noninvasively. It is a mathematical framework to generate stable locomotion with the help of a musculoskeletal prototype to control a robotic assistive device for walking.

In the April 2022 issue, a paper published in IEEE Robotics and Automation Letters outlined the AMBER team’s method and represented the first instance of combining hybrid zero dynamics (HZD).

Currently, HZD is employed to produce stable walking gaits for bipedal robots, and the muscle model illustrates how much a muscle extends or contracts with a given joint configuration.

Also ReadBat Bot Flying Robot – a bat inspired autonomous drone

The team is now working on its approach to demonstrate a battery-operated, motorized prosthetic leg. The battery controls the motors, which help to turn the joints. The mathematical algorithm directs the motor movement that the researchers develop.

While creating this mathematical algorithm, the AMBER research team recorded the muscle activity of an individual walking with a prosthesis that followed the desired motion instructed with HZD.

The process was conducted using electromyography (EMG), where one electrode is positioned on the skin above a specific muscle. The team then examined the EMG activity of the person with the prosthesis. The latter more closely resembles how a human makes movements without a prosthesis.

The team was keen to observe the muscle activity pattern of a human walking without the prosthetic. And by closely studying it and directly embedding musculoskeletal models into the optimization problem, the algorithm was ultimately able to produce gaits for the prosthesis, which is quite the foundation for generating gaits that can feel more natural.

The prosthetic device created by the research team has two actuated joints: the knee and ankle. Here, the knee and ankle joints follow their respective trajectories in response to the command sent to the motor. So, the desired motions and velocity of those joints over time is what the robot then performs. In this case, the robot is a leg prosthesis.

The incredible discovery the team came across was that the combination of HZD and the muscle models generated desired walking gaits, which were faster than they had been expecting.

Pushing the robotic model to pursue the patterns of muscle-tendon relationships adds further limitations to the gait-generation optimization problem, which is one problem that one might expect to be more challenging to solve. But with these additional conditions, a stable walking gait was developed after fewer iterations of the optimization problem.

The entire research and development by the AMBER team will help bridge the gap between techniques that use algorithms to produce desired walking motion and the field of biomechanics.

The resulting partnership brings Caltech’s AMBER lab team a step ahead in translating natural motion to a robotic assistive device like a prosthesis, with potential applications in full-body exoskeleton machines for people with paraplegia or other paralysis conditions.