SHENZHEN - Baidu launched Apollo World, an AI system that builds virtual road environments to train its Apollo Go robotaxis, at a Baidu AI Day session here on Tuesday, Reuters reported, as the company looks to cut the cost of preparing driverless fleets for rare hazards and new cities.

Baidu said Apollo World can generate up to 100,000 kilometers of simulated driving a day, which it compared with about three years of real-world testing by an autonomous vehicle carrying a safety driver, Reuters said. The system produces real-time camera and lidar data streams so driving software can sense and navigate the simulated scene, and it updates those feeds as the virtual vehicle steers, brakes or accelerates.

The company is using the simulator to rehearse hazards that are hard to capture safely on public roads, including cyclists cutting across intersections and animals running into a robotaxi's path, according to a South China Morning Post account of a demonstration by Yang Ze, who leads world models at Baidu's Intelligent Driving Group. Baidu also showed simulated trips in Shenzhen and Hong Kong with changing traffic levels, the Post said.

Baidu said Apollo World is already being used to train Apollo Go's robotaxi fleet for global operations. Apollo Go is China's largest robotaxi service and competes with WeRide, Pony.ai and Momenta Global, Reuters said. Baidu has said Apollo Go operates in 28 cities, has completed more than 23 million rides and has logged more than 350 million kilometers of autonomous driving, including more than 240 million kilometers with no one in the driver's seat, Gasgoo reported, citing the company.

Developing the technology remains expensive even when fleets are run with partners. "They still have the capex of training models ... and vehicles, a challenge that every autonomous vehicle company is going to face," Frank McCleary, an autonomous mobility expert at Arthur D. Little, told Reuters. Morgan Stanley has estimated China's robotaxi fleet could grow to 1 million vehicles by 2035, Reuters noted.