Shenzhen’s Chaowei Dynamics, a full-stack embodied-AI company, has entered a deep partnership with Peking University Medical to bring robots into clinical settings. The two will jointly develop data collection, world models and simulation training for embodied intelligence in medical scenarios — building, in their words, a path from “simulated training” to “scenario validation” to “hospital deployment.”
Getting robots into hospitals is harder than parking one in a ward. The bottleneck is well known in the industry: there is no environment to practise in, and no teacher to learn from — both realistic medical settings and high-quality clinical operation data are scarce. Chaowei brings the technical base; Peking University Medical brings the real clinical scenes and the expertise.
The first step is simulation. Chaowei’s KAI World Model engine recreates wards and operating rooms in a virtual world, shrinking the gap between simulation and reality so a robot can fail safely and iterate fast before it ever reaches a patient. A first-person data-capture headset, KAI Halo, records real medical staff at work and converts those operations into high-quality multimodal training data — effectively a roster of master trainers for the machine.
The second step is training inside that virtual hospital: an interactive digital asset library of drugs, instruments and equipment, on-demand scenario assembly, and large-scale operation learning that builds “muscle memory” through repetition. The third is validation and deployment — tube sorting, drug delivery and bed transport are the first scenarios earmarked for testing, chosen because they are the capillaries of a smart hospital and the safest places to prove reliability before anything riskier.
For a safety-critical setting like healthcare, the partners are deliberately moving slowly: get the simulated training solid, validate the scenarios thoroughly, then step into the hospital. No shortcuts, no overclaiming — just each capability earning its place in a real clinical environment.
*Translated and adapted from LeiPhone (https://www.leiphone.com/category/ai/xnczmPtqg4kZNirW.html).*