Team ExoMyo is developing a full body, robotic body suit for a paralyzed veteran competing in an international cybernetic competition, the 2024 Cybathlon. Recent strides in robotic motion, both in hardware and software have made more natural bipedal mobility a possibility. Leveraging existing and novel approaches, Team ExoMyo will partner mobility technologies and machine learning solutions along with a BCI control system to achieve balanced, efficient bipedal movement for a fully paralyzed individual. In this system, the motion intent will be directed by the human using BCI while balance and movement will be maintained by robotic motion control systems and algorithms. The suit itself will utilize the body’s existing skeleton and adds a soft shell “muscle” (myo in Greek) with electromagnetic actuators, not so much an “exoskeleton” as an “exomyo”.
What’s the focus of the project?
The ExoMyo design separates the control functions leaving the balance and power to a robotic system while intent is managed by the user through the BCI and muscle sensors. The BCI interface is used to train a machine learning dataset to learn intent (not individual limb control) and then used during piloting the suit to control direction and pace.
An “intention” model will be developed through testing on disabled and non-disabled participants using OpenBCI hardware and other sensors. We intend to initially test within established BCI paradigms (MI, SSVEP, etc) and train the data with a variety of machine learning models to evaluate and optimize performance efficacy, expanding beyond these BCI paradigms as necessary. While we anticipate the need for individual training of the Cybathlete, we believe that studying a larger population will provide insight into natural variance and aid in model selection and hyperparameter tuning.
Why is this work important?
Current exoskeletons are bulky and don’t often match the full degree of freedom for human bipedal movement. Making a lightweight, efficient, suit using BCI control will enable better agility and design success.
We will be looking to see how well a BCI system (and a machine learned reading of that data) will be able to read and interpret, with speed and precision, human intent. We will also see what other cues, like muscle groups if signals exist, in other parts of the body that aid in signaling intent of motion.
Who is involved in this project?
We are a team of design thinking, passionate engineers leveraging their backgrounds and knowledge in the design of the ExoMyo suit. An exciting part of this project is the opportunity to bring OpenBCI into the classroom. We will be incorporating some of this testing and research into a bioengineering course for high school juniors and seniors. These students will not only act as study participants but as active researchers making direct contributions. This work will introduce them to the world of BCI and, we hope, to inspire them to pursue it further.
- For further information: Please visit our website www.exomyo.tech or email [email protected] We will also periodically post our progress to the Community forum.
you say “current exoskeletons” but i don’t think i ever seen an exoskeleton become anything beside proof of concept… would this work ever become a real product? why would it be different from all the “current exoskeletons” that failed?
I’d like to acknowledge that a proof of concept isn’t a “fail” unless you’re talking about commercialization or going beyond the lab (which have many factors keeping it back). But I will say we are looking to address three fundamental challenges for current exoskeleton designs: agility, power and control. We are approaching the control issues (smooth, efficient and snappy response) with robotic and machine learning approaches. By letting the robotic system manage balance and individual actuators, then a BCI can handle intent which is much more likely than controlling individual muscle groups. Also, current exoskeletons have few degrees of freedom actuated by high profile servos. We are designing lower profile actuators that are placed more like the muscle groups are in the body. This design is enabled by a hard/soft body suit design that relies on the existing skeleton of the body. More “muscles”, more degrees of freedom and thus more agility. And finally the power… we are looking to make an efficient system relying on balance and the body weight with more actuators each using less power per unit.