At a conference in Shanghai’s Zhangjiang science park this month, a practical theme surfaced among the buzz around ‘embodied intelligence’: rather than chasing general-purpose humanoid robots, Chinese firms are targeting rehabilitation and eldercare as the first commercially and clinically tractable markets. Fourier, a Shanghai robotics firm, showcased a system that pairs non‑invasive brain–computer interfaces with exoskeletons and limb robots to close an "intention–execution–feedback" loop, turning patients’ motor imagery into assisted movement and continuous training data.
Rehabilitation medicine has structural problems that technology alone has struggled to fix: early severe patients cannot initiate movements, assessments are episodic and unquantified, and therapy is repetitive and demotivating. Brain–computer interfaces (BCIs) promise not to replace therapists but to restore a patient’s active role by detecting intent from EEG and other biosignals and translating it into device action, thereby harnessing neuroplasticity through synchronous brain and limb activation.
This vision did not appear overnight. Fourier began preclinical work in 2017 and has validated concept demonstrations such as EEG‑driven exoskeleton walking. Recent progress in lighter, multimodal BCI hardware and in AI decoding of neural signals has made practical trials more plausible. The company stresses hybrid sensor fusion—combining EEG, EMG and kinematics—and adaptive assistance that gradually reduces support as patients regain function, while using the multi‑dimensional data stream for continuous outcome measurement and personalization.
There are pragmatic reasons to start with care and rehabilitation. Demographic ageing in China is expanding long‑term demand for recovery, nursing and companionship services; these institutional settings tolerate richer human–robot interaction and provide safer, more controllable environments for validation than the unpredictable home. At the same time, core robot hardware—actuators, joints and motion controllers—is converging across suppliers, shifting competitive advantage toward system integration, data, and real‑world applications rather than chassis specifications.
Data scarcity is emerging as the industry’s bottleneck. Robot training demands first‑person, high‑quality interaction data that are expensive and slow to acquire; public videos address scale but not intention. Vendors are therefore combining multiple data sources and negotiating standardized interfaces. Fourier and several academic and medical partners, including Shanghai Jiao Tong University’s Ruijin Hospital and research institutes at Fudan and Tongji, have launched a joint ‘‘BCI‑embodied data engine’’ initiative to build hardware foundations, toolchains and shared datasets to accelerate safe deployment and model generalization.
Executives and researchers are cautious about timelines. Fourier’s founder and CEO, Gu Jie, expects technical convergence to become clearer in 2026–27 and predicts a staged diffusion from hospitals to institutions, community clinics and eventually homes. Observers see the next three to five years as a critical window: firms must translate lab breakthroughs into integrated products, clinical evidence and repeatable business models while forming ecosystems with hardware partners, hospitals and regulators.