Baidu launches Apollo World simulation model for robotaxi training

Baidu launched Apollo World on Nov. 3 and first disclosed it during an Apollo session at its AI Day in Shenzhen.[1][2] The company describes it as an agent-driven, multimodal world model for physical AI.[2] Its stated purpose is to simulate real-world environments for robotaxi training and predict how those environments evolve.[1] Baidu says the system is already being used to train Apollo Go’s robotaxi fleet for global operations; a separate report says it supports Apollo Go’s global testing and validation.[1][2]

Baidu says Apollo World can generate up to 100,000 kilometers of simulated driving per day.[1] The company compares that volume with about three years of real-world testing by an autonomous vehicle equipped with a safety driver.[1] The model can also produce real-time multimodal data streams, including camera and lidar observations.[1] According to the report, this simulated data lets self-driving systems directly sense and navigate virtual environments.[1] For teams assessing a training setup, the stated mileage and the type of sensor data are distinct measures to examine.

In Apollo World’s virtual cities, vehicles and pedestrians are modeled as AI agents with decision-making and interaction abilities, rather than as actors following fixed scripts.[3] The system supports driverless vehicles traveling between any starting point and destination in a virtual city.[2] It changes the virtual environment after a driving decision and feeds new observations back to the system, forming a closed loop.[2] Apollo’s proposed evaluation criteria include causality—predicting the next moment from the current state—and interactivity, where outside interventions can dynamically affect the world’s future state.[2]

Baidu-owned Apollo Go is described as China’s largest robotaxi service operator and competes with WeRide, Pony.ai and Momenta Global.[1] A separate report says Apollo Go had launched in 28 cities worldwide, completed more than 23 million rides and logged over 350 million autonomous-driving kilometers, including more than 240 million fully driverless kilometers.[2] Chinese robotaxi companies have generally pursued asset-light expansion through technology supply and partnerships, while autonomous-driving development remains expensive.[1] The reports describe Apollo World’s simulation capabilities and use, but do not establish how simulation results translate into performance on public roads.