Select Language:
China has officially opened its first fully self-developed robot simulation training facility in Wuhan’s Dongxihu district. This state-of-the-art field leverages breakthrough technology derived from a simulation platform created by a Beijing-based company specializing in embodied intelligence.
One of the primary hurdles in embodied intelligence is the scarcity of data. Gathering real-world data involves setting up physical environments, deploying numerous robots, teleoperation equipment, and personnel, which makes data collection both inefficient and expensive. This also limits scenario diversity, resulting in high-quality data output that is nonetheless insufficient. To address this, virtual environments are used to generate simulation data, solving cost and scenario limitations. However, creating virtual worlds that accurately mimic real-life settings remains a challenging task.
Simulation training acts as a “virtual school” for robots. From shopping in supermarkets and serving in restaurants to home assistance and industrial tasks, a variety of highly realistic scenarios continuously produce training data, laying the groundwork to enhance robots’ “brains,” according to industry experts.
Participants can wear virtual reality headsets and immerse themselves in digital environments designed to teach robots. For instance, during a demonstration, a reporter picked a drink from a virtual supermarket shelf, and a real robot immediately recognized the scenario, retrieving the same item. Additionally, virtual tasks like folding a quilt resulted in real robots executing identical actions neatly.
The core computing engine powering this simulation platform uses significantly less energy—less than a tenth—compared to comparable foreign products. It is compatible with domestic chips and operating systems and boasts industrial-grade precision. Its training scale, accuracy, and computational efficiency surpass those of leading international companies, although there remain some gaps in ecosystem development.
The developers emphasize a commitment to innovation from the ground up, with every component—code, tools, and data pipelines—entirely self-created. This has enabled complete technical independence and control over the platform.
Several prominent robot manufacturers, including Unitree Robotics and Xinghaitu, are already integrating this platform for trial training sessions.
ChatGPT
Perplexity
Gemini AI
Grok AI
