X Square Robot has introduced Wall-B, a physically embodied artificial intelligence prototype designed for household use, and said it plans to begin deploying robots in private homes within 35 days. The Beijing-based company also presented its World Unified Model architecture, a training framework intended to integrate image processing, speech, motion and physical prediction into a single system. X Square Robot said the model is designed to help robots operate in household environments, where layouts, tasks and interactions can vary widely.
“Robots in factories and robots in private homes are fundamentally different,” said Qian Wang, founder and CEO of X Square Robot. “In factories, they repeat the same action 10,000 times. In a private home, they may have to perform 10,000 different actions, each in a different context. The real challenge is not the repetition, but whether a robot can perform new, untrained actions in an unstructured environment.”
Wall-B is the first full implementation of the company’s World Unified Model architecture, according to X Square Robot. The company said the approach trains perception, speech, control and physical prediction together, rather than as separate modules, with the aim of allowing the system to anticipate force, friction and collision dynamics.
“We train seeing, speaking, acting, and predicting from the very beginning in the same network,” said Wang Hao, chief technology officer of X Square Robot. “Human infants don’t learn to see, move, and communicate in isolated phases. They learn by integrating perception and action simultaneously, with constant feedback from the physical world. That’s the principle behind our architecture.”
The company said the model is built around data from real home environments and a physics-aware predictive mechanism designed to help the robot anticipate physical outcomes before acting. X Square Robot said these elements are intended to address the challenge of moving from controlled demonstrations to household use.
At the launch event, the company demonstrated tasks including flower arrangement, with a robot adjusting its grip and movements as stems shifted under visual occlusion. X Square Robot said the task was performed without preset movement paths.
The company acknowledged that the technology remains at an early stage. Qian Wang said current systems can still make errors requiring remote intervention, including leaving objects in the wrong location or pausing during tasks while processing the next action. X Square Robot said its planned household deployments are intended to generate real-world operating data as it continues developing household robotics.
