Baidu Unveils 'Apollo World' AI Simulator for Robotaxi Training

At a glance

Chinese company Baidu has unveiled Apollo World, an AI simulator that trains its robotaxis on virtual roads. The goal is to improve self-driving performance by creating dangerous situations that are hard to encounter on real roads.

Key points

1. 1

   When a vehicle responds to its steering wheel, brakes, and accelerator, camera and lidar data are generated immediately.

2. 2

   It recreates rare dangerous situations from real roads in a virtual setting, such as a cyclist running a red light or an animal darting into the road.

3. 3

   Previously, traffic rules had to be coded by hand for each city, but the system introduces a learning method that turns sensor input directly into driving commands.

4. 4

   Baidu is preparing to expand overseas, including cross-border driving between Shenzhen and Hong Kong.

Why it matters

For self-driving cars to operate safely across many cities, they need enough practice with rare dangers, and this technology is changing how that practice happens.