Make walking robotic assistant more natural

Make walking robotic assistant more natural

A team of graduate students in the California Institute of Technology’s Advanced Mechanical Biped Experimental Robotics Laboratory (AMBER), led by Professor Aaron Ames, Bren Professor of Mechanical and Civil Engineering and Dynamic Systems and Control, is developing a new method for gait generation for robotic assistive devices, which aims to ensure stability And achieve more natural movement for different users.

a paper Published in the April 2022 issue of IEEE Robotics and Automation Letters It summarizes the AMBER team’s method and represents the first example of combining hybrid zero dynamics (HZD) – a mathematical framework for generating stable movement – with a musculoskeletal model to control a walking assistive robotic device. The musculoskeletal model is a computational tool for measuring the relationship between muscle strength and joint strength. HZD is currently used to create a stable gait for bipedal robots, and the muscle model represents how much a muscle is stretched or contracted with a given joint configuration.

The team demonstrates their approach in a battery-powered, motorized prosthetic leg. The battery powers the motors that run the joints. The movement of the motor was determined by the mathematical algorithm developed by the researchers.

To create this mathematical algorithm, the AMBER research team recorded the muscle activity of a walking person using a prosthesis that followed the desired movement generated using HZD alone. This was done using electromyography (EMG), in which a single electrode is placed on the skin over a specific muscle. The team then analyzed the EMG activity of a walking person with a prosthesis that followed the desired movement generated by HZD along with muscle models. The latter is very similar to how a human walks without a prosthesis.

“The pattern of muscle activity of a human walking without a prosthesis is what we want to get close to,” says Rachel Jahler (MS ’18), a graduate student in mechanical and civil engineering. The direct inclusion of musculoskeletal models in the optimization problem—the algorithm that ultimately produces gaits for the prosthesis—provides a basis for generating a gait that can appear more natural.

“If you are designing a pathway for an automated assistive device, not only should the pathological gait be stable, but it will also feel normal,” says Amy Lee, a graduate student in Computing and Neurological Systems. The trajectory represents how researchers want the prosthesis to move over time.

“One way to control the robot is to describe the desired movements of each joint,” adds Maegan Tucker (MS ’19), a graduate student in mechanical and civil engineering. The robot in this case is a leg prosthesis. The knee and ankle joints follow their own paths in response to the command sent to the motor. “Our prosthetic device has two stimulating joints: the knee and the ankle. So the desired movements and speed of those joints over time is what we’re asking the robot to do.”

One of the surprising discoveries was that the combination of HZD and muscle models generated desirable gait faster than expected. Forcing the automated model to follow patterns of muscle-tendon relationships adds further limitations to the gait improvement problem, so one might expect the problem to be more difficult to solve. But with these additional limitations, a stable walking gait was developed after less repetition of the improvement problem.

“Our assumption is that the built-in model helps direct problem optimization toward solutions, which lead to normal walking,” Tucker says. “In other words, the additional constraints help direct the optimization problem toward a more stable and natural gait faster.”

This work helps bridge the gap between methods that use algorithms to produce desired gait motion and the biomechanical field, which usually do not overlap. The resulting collaboration leads the AMBER lab a step closer to translating natural movement into an automated assistive device such as a prosthesis, with potential applications in full-body exoskeleton devices for people with paraplegia.

View a . file View a video For the auxiliary device described in this story.


#walking #robotic #assistant #natural

Leave a Comment

Your email address will not be published. Required fields are marked *