Embodied AI Advances at Beijing Robotics and Brain Science Events

At the WRC, which opened Aug. 19, Xingchi Power displayed a robot platform using PULSE to handle different scenarios. The company defines PULSE as a user-intent-driven world-action model that connects perception, intent understanding, reasoning and simulation, action execution, and feedback evolution. The system runs on a cloud-edge-device architecture, with the edge responsible for real-time sensing and low-latency decisions, and the cloud for global optimization. PULSE adapts to different robot bodies, allowing capabilities to be reused across products.

Xingchi Power also exhibited first-person-view data-collection devices that capture human demonstrations. However, the company said first-person data only begin the data chain; true-robot data—including joint structure, sensor configuration, motion range, control latency, and failure and recovery processes—are needed to calibrate and verify actions. The firm calls this the 'last 100 meters' of data collection and plans to feed execution results, failure cases, manual corrections, and user feedback back into PULSE for iterative improvement.

The WRC demonstrations followed the 2026 Beijing Brain Science International Conference, held Aug. 14–16 and co-organized by the Beijing Brain Science and Brain-like Research Institute (CIBR) and Nature Portfolio. The BRIGHT Foundation presented an academic poster titled 'Beyond Top-Down Control: Operationalizing Connectome Circuits for HIE Restoration via The NeuroLoop Protocol.' On Aug. 17, it convened a workshop with ten experts from areas such as basic neuroscience, engineering, AI, and clinical practice. Participants included researchers from CIBR, the University of California, San Diego (including Don Cleveland), the Chinese University of Hong Kong, the University of Göttingen, and clinicians from Peking University International Hospital.

Discussions ranged from the role of sensory feedback and motor control in recovery to non-invasive brain stimulation and wearable systems. One expert noted that many everyday movements rely on subcortical and spinal closed loops, and another highlighted the potential of low-cost sensors and markerless motion capture for behavior analysis. A clinician stressed that careful evaluation is key to avoiding unnecessary surgery, and that families need both hope and certainty.

According to Leiphone.com, BRIGHT Foundation founder Matt Palaszynski said that cerebral palsy and chronic brain injury have long been overlooked by mainstream brain-computer-interface investment because their distributed, non-linear trajectories do not fit traditional financial models. He argued that China’s push for 'new quality productive forces' could reframe these conditions as opportunities for public health and macroeconomic value. He also said NeuroLoop is an open conceptual framework, not a product, and that the foundation is using an AI-driven intelligence platform to screen research literature and build individualized 'N-of-1' predictive models.

Palaszynski noted that near-term engineering opportunities may lie in combining non-invasive stimulation, such as transcutaneous auricular vagus nerve stimulation, with task-oriented soft robotics and wearables. He said the foundation is collaborating with Peking University International Hospital, CIBR, Tsinghua University, and Zhongguancun College. He added that the main bottleneck is not hardware or software capability but system integration and translation, and that an organizer is needed to push the transformation ecosystem forward.

At the WRC booth, visitors watched the robot respond to sudden instructions, an interaction that suggested a growing industry focus on adaptability in real-world environments. The company said the exhibition was a starting point, with the real world providing ongoing answers.