KEYNOTE Google's Self Driving Cars
summary
Google builds a self-driving Prius with a spinning LiDAR that produces 100,000 3D points per rotation at 10 Hz (≈5 cm accuracy), long-range radar, cameras, drive-by-wire embedded controllers and a workstation running lightly modified Ubuntu with SCHED_FIFO and cgroups. Digital maps use 10 cm pixels with infrared and height; GPS (5–30 m) is refined by laser-to-map alignment. The fleet logs and replays 400,000 miles of autonomous driving; it beats employees on a test track, struggles in heavy rain/snow and hands control to a human for rare failures.
AI-written summary · about
chapters (12)
- 0:00Start
- 4:16Why Is Google Even Qualified To Do this
- 4:42Approaches to Date for Self-Driving Cars
- 6:45Sensors
- 7:33Radar
- 8:07Cameras
- 8:38Positioning System
- 8:50Computing Platform
- 9:32How We Use Linux
- 13:48Digital Maps
- 15:27Maps in 3d
- 21:19How Do You Deal with Snow
related talks
21mThe race to replace human drivers
35mDingoCar: Perception module Small scale, self-driving car
31mAutonomous driving, how Python drives Formula Racing cars
26mSelf Driving cars with Opencv
22mA Fully Immersive Driving Simulator for Autonomous Vehicles
28mMachine Learning Inside the Car
35mGetting started with a self-driving RC car
46mKeynote: Make your car self-driving using open-source software