Autonomous Driving

Letting a car perceive road conditions, make decisions, and control itself with little or no human input.

Autonomous driving means a vehicle uses sensors such as cameras, lidar, and millimeter-wave radar, together with an onboard computer, to handle perception, decision-making, planning, and control on its own. SAE J3016, the standard from the Society of Automotive Engineers, defines six levels, L0 through L5. At L2, the system handles both steering and speed, but the driver must watch the road at all times. From L3 up, the system takes over driving under specific conditions and asks the driver to step in when needed. L4 allows fully driverless operation within a defined area, and L5 needs no human under any condition. In China, “智驾” (“smart driving”), the everyday term, usually refers to L2-level driver assistance. Autonomous driving can be seen as the first form of embodied AI to reach large-scale deployment, and it shares techniques with robotics more broadly, including end-to-end models, world models, simulation, and data flywheels — the industry often talks about “crossing over from autonomous driving into embodied AI.”

ExampleUrban NOA (navigate-on-autopilot) can automatically follow traffic, change lanes, and go through intersections in a city, but the driver must stay ready to take over at any time — this is an L2-level system.

Also called

Self-driving

Related

End-to-End · World Model · LiDAR · Levels of Autonomy · Autonomous-Driving Talent Moving into Embodied AI · Embodied AI

Sources

Self-driving car - Wikipedia（含 SAE J3016 分级） (Chinese)