The $546,848 award funding the project over the next five years will include collaborations with medical schools, rehabilitation centers, and the robotics industry in Rhode Island and Massachusetts. (Abiri encourages people with disabilities who’d like to become involved with the project to contact him at Reza_Abiri@uri.edu.)
Abiri is also using the award to help inspire the next generation of scientists, with a new university course, annual hack-a-thon competitions, and internships to pique the interest of K-12 students in human-centered robotics and AI.
Abiri joined URI’s electrical, computer and biomedical department in 2021, and started his research laboratory on Translational Neurorobotics.
This project will seek to connect a person’s small movements, like a head turn or eye movement, to something that assistive robotic arms using artificial intelligence can understand.
What it could mean for people with spinal cord injuries is the ability to gain independence through artificial intelligence and robotics.
Q: Why does the field of translational neurorobotics interest you?
Abiri: We’ve already spent so much money for research for the past 20, 30 years, especially in neuroscience research, and right now, it’s the time to translate how these kind of sciences can be used for people.
For stroke patients or those with spinal cord injuries, the most important thing is how we can help with patients with motor impairment. Usually when you talk about stroke, you expect to see some recovery. But for the case of a spinal cord injury, usually it’s very hard to imagine. You need some assistive technology to help them be independent in their activity.
How do you see AI working in this?
Our work is how the human and AI can collaborate to complete a task. You have heard a lot about Chat GPT, but that is a text generator, so it doesn’t necessarily have an understanding of how a human is moving. We are trying to have a completely different approach for how AI can understand human movement, and how you can project that into a robotic system.
For example, my hand goes right, my hand goes left, but the point is how AI can understand and move a robotic system the way that the human wants.
How do you see this working in the world, when you’re able to start testing it on some of the patients who have spinal cord injuries?
The way that we are working for this AI pipeline is that hopefully it can understand the human input. That means it can be a brain activation or a muscle activation, or even a very simple head motion. So, it is just the matter that the AI can capture those signals from the human and say, OK, I understand this.
Say that my intention is to move that robotic arm to the right and left, but it’s not exactly all the information you need to move. Maybe on the right there is a cup of coffee. To grab a cup of coffee, the AI needs to understand how to do that.
What the AI is doing is, ‘I understand that you want to move to the right, and what I see on the right’ — because it also has a camera — ‘is a cup of coffee. I know that a cup of coffee is not supposed to be grabbed from the top.’ It’s supposed to grab from the side and then bring it up. It doesn’t necessarily match up all the information that the human is giving, but we are trying to make a smarter AI to understand all this motion from the human.
So, you’re not only meeting the patient where they’re at and where their capabilities are, but also using AI to understand? It isn’t just that I want you to move — it’s what I want you to do when you move, whether it’s to pick up a pen, or a cup of coffee, or pet a dog.
Yes, exactly. And one of the interesting goals behind this is that this system should understand the individual’s capability. Each individual is different. Each individual has their own disability, their own way of interacting with the environment. If we have a better understanding of this human subject and how they perform some limited task, then we can translate those to the real environmental interaction to do complex tasks.
I think this trend of human-AI collaboration is the next stage of work. Right now, if you have Chat GPT, it is just one-sided. The human is not doing anything. I hope that this foundation can create a bigger region of human-AI collaboration, and then AI could be an assistant beside the human.
You’ve worked with human subjects before. How do they respond when they become involved with a project like this?
They’re excited to see this technology, and that they are really allowed to do something on their own. Usually in most cases these individuals with motor impairment prefer to be involved, rather than to have someone just perform something in an autonomous way.
You’re kind of giving their independence back, where they get to decide what they want to do.
Yes. And then with some sort of boosting mechanism that can help them, we are trying to make this a more interactive pipeline, where the users become more motivated to use their body to interact with the environment.
So, you can imagine, that if I wanted that person to get engaged with the robot interaction or the computer interaction, and then they become more interested —maybe try this movement, maybe try that moment — then this pipeline will learn and then kind of step back. Maybe that pipeline understands that right now this person can do that, and I should step back and just be another part of the task. We hope that this patient becomes much more excited about this interaction, and they can engage with this system
In a way they’re also training AI on how to be more useful as well, aren’t they?
Exactly. And what we hopefully can imagine is this could be a personalized and adaptive AI that can work for that specific individual. For each individual, they have their own capabilities. Hopefully, this pipeline can be useful for them to do more personalized ways of training the AI for themselves.
So the impossible question is, how far away do you think we are from something like this?
I think from the perspective of autonomous, we are in a good stage. But from the perspective of human involvement, I see this is the beginning.
The Boston Globe’s weekly Ocean State Innovators column features a Q&A with Rhode Island innovators who are starting new businesses and nonprofits, conducting groundbreaking research, and reshaping the state’s economy. Send tips and suggestions to reporter Alexa Gagosz at alexa.gagosz@globe.com.
Amanda Milkovits can be reached at amanda.milkovits@globe.com. Follow her @AmandaMilkovits.
link