Baidu Unveils New AI System to Test Self-Driving Cars

Chinese technology firm Baidu launched an artificial intelligence system named Apollo World on Tuesday to simulate real-world road environments for robotaxis. The new model can generate up to 100,000 kilometres of simulated driving every single day. Baidu states this daily output equals about three years of real-world testing using a safety-driver-equipped autonomous vehicle.

Fleet Operations

The system produces real-time data streams including camera and lidar observations. This allows self-driving systems to directly sense and navigate the virtual environments. Baidu is already using Apollo World to train its Apollo Go robotaxi fleet for global operations. Apollo Go currently stands as China's largest robotaxi service operator and competes with rivals such as WeRide, Pony.ai and Momenta Global. Meanwhile, Volkswagen Chinese partner XPeng plans to launch its own robotaxi operations in 2027.

Market Growth

Financial firm Morgan Stanley estimates that China's robotaxi fleet could grow to one million vehicles by 2035. Most Chinese robotaxi firms pursue asset-light expansion strategies. They supply self-driving tech and partner with local fleet operators and ride-hailing platforms overseas rather than paying the full cost of owning fleets. However, developing the underlying technology remains expensive. Frank McCleary, an expert in autonomous mobility at Arthur D. Little, notes that companies still face capital expenditure challenges for training models and vehicles. Tesla also works on similar technology, using a neural network trained on human driving videos for its Cybercab fleet.

Context

Simulators help autonomous vehicle developers test rare road scenarios safely without risking physical cars or people. Lowering the high cost of training self-driving software remains a central goal for the entire global industry. Major technology firms continue to invest heavily in artificial intelligence models to speed up vehicle development.