Loading page…
Baidu claims Apollo World can generate a day’s simulated driving equal to about three years of testing by one safety-driver-equipped vehicle.
Listen to this story
Baidu says Apollo World is already training Apollo Go for global operations, making the simulator part of the robotaxi program’s current development pipeline rather than a future-only tool. It generates changing virtual driving environments with real-time camera and lidar data, which Baidu says amounts to about three years of testing by one safety-driver-equipped vehicle each day. The system could help address the cost and time of learning new operating environments, but the testing-volume comparison is Baidu’s own claim.
Baidu launched Apollo World on November 3, 2026, to simulate roads and predict how their surroundings change.
Apollo World supplies sensor observations, including camera and lidar data, for autonomous systems to navigate simulated environments.
Fleet partnerships can reduce robotaxi operators’ need to own vehicles, but they do not remove the capital required to develop and train driving technology.
Baidu launched Apollo World on November 3, 2026, giving its robotaxi training program an AI system designed to recreate driving environments and predict how they change. The launch targets an expensive part of autonomous driving: developing technology that can learn to navigate the places where a service will operate.
Chinese robotaxi companies have largely pursued expansion through partnerships rather than paying the full cost of owning and running fleets. They supply autonomous-driving technology and work with automakers, ride-hailing platforms and local fleet operators overseas. That approach does not remove the expense of developing the driving technology itself.
Frank McCleary, an autonomous-mobility expert at consultancy Arthur D. Little, told Reuters that model training and vehicles still require capital investment. The speed of a launch, he said, depends on how well the technology can adapt to and learn a driving environment.
Apollo World’s role is to supply environments in which self-driving systems can practice. It is designed both to simulate real-world surroundings and to predict how those surroundings evolve, making changes in the environment part of the training task.
Baidu says the system produces real-time streams of different kinds of sensor data, including camera and lidar observations. Those streams let autonomous-driving systems directly sense and navigate the simulated environment. The output is therefore not just a road scene to view: it supplies observations that a driving system can use while moving through that scene.
The system is not being presented solely as a future training tool. Baidu says Apollo World is already being used to train Apollo Go’s robotaxi fleet for global operation. Apollo Go, owned by Baidu, is China’s largest robotaxi service operator and competes with WeRide, Pony.ai and Momenta Global.
Loading discussion...
Join the conversation
Explain how much weight you give virtual practice versus real-road experience.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.