Baidu’s Apollo World Simulates 100,000km Daily for Robotaxi Training

• Apollo World generates virtual environments from real-world sensor data to safely accelerate self-driving development
• Platform sets new benchmarks for China’s rapidly expanding robotaxi market, simulating three years of testing in a single day

On November 4, 2026, Reuters reported that Baidu introduced Apollo World, an artificial intelligence simulation platform capable of generating virtual driving environments for its Apollo Go robotaxis. The platform simulates up to 100,000 kilometers of driving per day—equivalent to three years of real-world testing by a single vehicle—enabling Baidu to speed up autonomous vehicle development and lower operational costs for robotaxi operators.

Apollo World uses actual sensor feeds, such as camera and lidar data, to create more realistic and dynamic scenarios. This allows self-driving cars to encounter rare hazards and complex road conditions that seldom occur in real traffic, improving the safety and readiness of autonomous vehicles in unpredictable situations.

Baidu has begun deploying Apollo World for global expansion efforts of Apollo Go, joining a lively race with technology firms such as WeRide, Pony.ai, and Momenta Global. Designed for scalability, Apollo World introduces Baidu’s “2+3” testing framework, a new industry standard for world model validation in autonomous vehicle training.

The launch comes as China’s robotaxi sector grows rapidly. Morgan Stanley estimates the market could support up to 1 million robotaxis by 2035. As competition intensifies, firms like XPeng, partnered with Volkswagen, plan to launch their own robotaxi services by 2027, driving innovation across the industry.

Beyond autonomous vehicles, Apollo World’s modular architecture enables simulation for broader robotics applications, reflecting Baidu’s push to lead in physical AI technologies.