The hardest lessons for self-driving cars
According to the South China Morning Post, the hardest lessons for autonomous vehicles are the situations that test even experienced human drivers: a cyclist darting through a red light or an animal bolting into traffic from behind a corner. Baidu developed Apollo World to prepare its robotaxis for such rare but high-stakes hazards, as it seeks an edge in the fast-growing robotaxi market.
On Tuesday, Baidu's Intelligent Driving Group showcased the new “world model,” showing how the simulator reacts in real time to the vehicle's steering, braking and acceleration while generating synchronized camera images and lidar data. The system can build complete virtual journeys from scratch, preparing vehicles for scenarios that are hard to capture safely in real-world testing.
We hope our world model can evolve from simulating a single scenario to simulating the execution of an entire task.
Stress-testing dangerous scenarios
In technical demonstrations, Yang showed Apollo World generating high-risk hazards such as cyclists cutting across intersections and dogs running into a robotaxi's path. Engineers can tweak the behavior of virtual road users to stress-test the driving software's responses under varied conditions.
















