WALL‑B: home‑robot foundation model
- X Square Robot said on April 22 it launched WALL-B, a home-robot foundation model, and plans to place its next generation of household robots in real homes within 35 days. - The company said WALL-B trains vision, language, action and physical prediction together, using data from more than 100 households to handle chores in messy, changing home environments. - The push comes as Chinese robot firms shift from stage demos toward domestic work, where slow, error-prone manipulation remains the hardest test. (independent.co.uk)
A home robot needs more than balance and motors. It has to see clutter, understand instructions, and handle objects that move, slip, bend, and break. (prnewswire.com) (independent.co.uk) That is the problem X Square Robot said it is tackling with WALL-B, an embodied artificial intelligence model unveiled in Beijing on April 22. The company said robots using the system will begin entering households within 35 days. (prnewswire.com) (marketwatch.com) Most robot systems split the job into separate parts: one model for perception, one for language, one for control. X Square said WALL-B instead trains vision, language, action, and physical prediction in one network from the start through what it calls a World Unified Model. (prnewswire.com) (en.yna.co.kr) Physical prediction is the part that tries to guess what happens next in the real world: how much force a grip needs, whether an object will slip, or where a collision might happen. X Square said that is necessary in homes, where lighting, layouts, and tasks can change from room to room and minute to minute. (en.yna.co.kr) (prnewswire.com) Chief executive Qian Wang said factory robots repeat the same action thousands of times, while home robots may face thousands of different actions in different contexts. Chief technology officer Wang Hao said the design goal was to let perception and movement learn together, rather than in isolated stages. (prnewswire.com) (en.yna.co.kr) The company said WALL-B was trained with simulation and real-world household data, including material collected from more than 100 homes. Reuters, cited by other outlets covering the launch, reported the public demo focused on slow, basic tasks such as picking up litter and arranging flowers. (robbreport.com) (independent.co.uk) That choice was deliberate. Folding clothes, loading a dishwasher, or grasping small objects is harder for robots than running or dancing because tiny errors in position or force can ruin the task. (independent.co.uk) (robbreport.com) X Square is not presenting this as a lab-only project. The company said the next hardware generation tied to WALL-B will be sent into homes in late May, and it framed those placements as part of learning in real domestic settings. (prnewswire.com) (gasgoo.com) The financing behind that push is substantial. China Daily, cited by The AI Insider, said X Square recently raised nearly $276 million in a Series B round led by Xiaomi’s strategic investment arm, adding to backing from Alibaba, ByteDance, Meituan, and Xiaomi. (theaiinsider.tech) (marketwatch.com) The broader race is shifting from robots that impress on stage to robots that can survive a kitchen, bedroom, or laundry room. WALL-B’s test is not whether it can perform a demo once, but whether it can keep doing ordinary chores in an ordinary home. (independent.co.uk) (eweek.com)
