Researchers at the University of Chicago have built a wearable system that uses artificial intelligence together with electrical muscle stimulation (EMS) to physically guide a user’s muscles, allowing the person to perform correct movements by being led through them rather than only receiving verbal or visual instructions.
What the system consists of
The prototype combines four main components:
- an electrode-equipped wearable suit (driven by EMS),
- smart glasses with a built-in camera,
- a motion-tracking layer to sense body posture in real time,
- a multimodal AI model capable of processing both visual and linguistic inputs.
Rather than playing back fixed motion patterns, the system uses camera and motion-tracking data so the AI can generate on-the-fly sequences of electrical impulses tailored to the current task, targeting specific muscles and joints.
How teaching works in practice
The team gives an example: if a user approaches an unfamiliar window and says, “EMS, help me open this,” the glasses identify the type of handle. The model then determines which fingers, wrist and elbow movements are required and electrically guides the appropriate joints through the correct sequence.
Safety and adaptability
The novelty lies in the AI’s ability to adapt in real time to the environment and the user’s bodily state. The researchers built an anatomical safety layer between the AI and the body: if the model would command a movement that could cause injury (for example, a 180-degree wrist rotation), the control system redistributes the motion across multiple joints to avoid a dangerous posture.
In lab tests, the full system made significantly fewer errors than a basic AI model that lacked this body-awareness layer. When the researchers intentionally introduced errors during testing, users typically noticed, adapted, and corrected the motion by reissuing commands or otherwise compensating.
Limitations and future directions
The team acknowledges several current limitations. Electrodes must be calibrated for each person, the electrical stimulation can sometimes be uncomfortable, and the system does not yet produce true long-term muscle memory — the kind of deep, practiced skill that comes from repeated training.
Nevertheless, the researchers note that both AI and EMS hardware are advancing quickly. They suggest that body-worn, AI-controlled muscle actuators could become as common in the not-too-distant future as today’s wearable health devices.
Potential applications
Possible uses include teaching physically complex skills (for instance, industrial motion sequences or musical instrument techniques) and assisting people with situational disabilities. Project lead Pedro Lopes has highlighted that this approach could represent a breakthrough for both skill training and assistive technologies.
Conclusion
The University of Chicago’s system integrates vision, language understanding and EMS to deliver real-time, context-aware muscle guidance. While practical challenges remain — per-user calibration, discomfort from stimulation, and the absence of true muscle memory formation — the researchers argue that rapid progress in both AI and EMS hardware could make similar wearable actuators widely available in the future.
