Intel and Onyx Healthcare bring AI-powered exoskeleton rehab to more hospital clinics

Intel says its technology is now helping power an AI-driven medical exoskeleton designed to speed and standardize rehabilitation for people with lower-limb injuries that leave them unable to stand or walk. The company highlighted the work in a new update describing how Onyx Healthcare, a Taiwan-based medical device manufacturer, has built a wearable system that guides patients through repetitive, limb-focused movements meant to support neuroplasticity, the brain’s ability to relearn motor skills.

The effort centers on FREE Walk, a robotics-based AI rehabilitation solution intended for clinical training sessions. According to Intel, the goal is not only improved mobility in the near term, but eventually helping patients walk again. The company contrasted the approach with traditional rehabilitation, which it characterizes as fatigue-inducing, imprecise, and dependent on hours of hands-on supervision from expert therapists.

Intel’s role, as described in the announcement, is to supply compute and software used inside the exoskeleton. The company says the system uses Intel Core Ultra CPUs along with Intel OpenVINO, a software toolkit for deploying AI models across hardware types. Intel’s announcement also describes how the AI runs on-device, aiming to support real-time guidance during sessions rather than relying solely on off-device processing.

On the sensing and control side, Intel says algorithms running on the Core Ultra processor work alongside pressure sensors, electromyography (EMG) force sensors, and inertial sensors. EMG refers to measuring electrical activity related to muscle activation. Intel says these indicates are translated into predictions about the patient’s next level of movement, with the system targeting gait stability by combining bio-electric muscle indicates with movement tracking.

Onyx Healthcare, founded in 2010 and headquartered in Taiwan, is described as the medical device maker behind the exoskeleton platform and clinical use. In Intel’s account, the company worked with Intel and FREE Bionics to design the AI-powered exoskeleton robot, with an emphasis on making rehabilitation sessions safer and more consistent. Intel says OpenVINO is used to help engineers evaluate and deploy the right AI models across CPU, neural processing unit (NPU) and GPU resources, with the intention of reducing latency and maximizing performance at the “edge,” meaning directly on the wearable device.

The announcement also offers a glimpse of scale. Intel says FREE Walk exoskeletons are used by approximately a third of medical-center-level hospitals in Taiwan. It adds that the system has accumulated more than 40,000 patient training records across 22 other countries. Those details matter for the companies because larger datasets can support iterative improvements to AI models and the broader deployment of an AI-enabled medical device into routine clinical workflows.

Intel further argues the system can ease staffing constraints in rehabilitation. In traditional lower-limb rehabilitation, Intel says a single session may require up to four therapists to assist one patient. With the AI-powered exoskeleton, Intel states therapist support can be reduced by at least two people, which it characterizes as improving staffing efficiency by about 50% per session. Intel ties this to the exoskeleton’s ability to guide patients through safe, consistent and repetitive movements while shifting some responsibilities away from multiple therapists.

From an engineering standpoint, Intel says the platform keeps time-sensitive AI processing close to the device during training sessions, then uses accumulated training data for cloud-based review and product improvement. That combination is framed as a way to balance real-time responsiveness, which is important in physical training environments, with longer-term learning and scaling of the system as more data is gathered.

As with many medical AI and robotics platforms, the announcement leaves important clinical and regulatory specifics unstated. Intel does not provide outcome metrics such as changes in walking speed, therapist-assessed gait improvements, or quantified functional gains, nor does it describe the regulatory status of FREE Walk in each country. Intel also notes general performance qualifications in its release, including that performance results depend on configuration and testing dates and may not reflect all updates, and that it does not control third-party data.