Over the last few years our BLINCdev team has been working hard to develop a battery powered onboard controller to allow our Bento Arm platform (5 degree of freedom robotic arm built with Dynamixel MX series actuators and controlled through our brachI/Oplexus software via the Dynamixel SDK) to be used as a fully portable upper limb prosthesis in addition to the desktop applications that we have traditionally supported. In October 2024, we put our newly developed Bento Arm Prosthesis V1 to the test in the Arm Prosthesis Race of the Cybathlon, a global competition where pilots of assistive devices compete against each other in tasks similar to ones that persons with disabilities encounter in their daily lives. We finished 8th overall out of 18 teams and were the top finisher from North America. We recently released a mini documentary telling the story of our team’s development and some highlights from the competition:
We are pleased to report that the Dynamixels actuators that we used in the prosthesis were super reliable throughout our development and competition cycle. Many other teams had hardware failures leading up to and during the competition, but our hardware ran without issue even when we ended up having to compete in cooler temperatures than we expected (8-12 degC compared to most of the teams which competed in 18-20degC temperatures) The relative low weights of the actuators combined with 3D printed links also allowed us to get better performance out of the machine learning controller which maps from the pilot’s muscle signals to the elbow and wrist of the prosthesis. Using actuators with built in sensors that communicate over a digital bus also gave us more flexibility to adjust settings on the prosthesis on the fly allowing us to optimize the range of motion, sensitivities, and speeds of the prosthesis joints to each individual task in the competition.
We are hoping to release the open source files for the Bento Arm Prosthesis V1 later this year including all the details of how we got it working with the battery system (which may also be helpful for other projects using Dynamixels). In the next iteration of our competition prosthesis we will focus on developing improved machine learning controllers and transitioning towards using Dynamixel X series actuators.