Autonomous Driving Levels
SAE 自动驾驶分级
SAE,美国汽车工程师协会
L0 无自动化
L1 提醒驾驶
- 加速、转向、制动等环节由人类自己操作
- 系统只提供提示、预警类信息
L2 干预驾驶
- 部分自动化
- 基本驾驶操作仍由人类完成,但当系统感知到危险时会进行短暂的驾驶干预来保障驾驶安全
- e.g. 自动变道,车道偏离预警
L3 人机共驾
- 有条件自动化
- 部分驾驶操作可由自动驾驶系统完成,如系统可在部分场景下控制加速、制动及转向操作,非完全自动驾驶
L4 辅助自动驾驶
- 高度自动化
- 由无人驾驶系统完成全部驾驶操作
- 人类需要根据系统请求接管车辆
- 非完全自动驾驶,需要限定驾驶环境和条件
L5 完全自动驾驶
- 完全自动化
- 由无人驾驶系统完成全部驾驶操作
- 人类可以根据需要接管
- 在所有道路和环境条件下均可进行自动驾驶
自动驾驶任务分类
一般来说,Driving Task 可以分为三个部分:
- Perceiving the enviroment
- Planning how to reach from A to B
- Controlling the vehicle
ODD, Operational Design Domain. 限定操作域。
具体驾驶任务由哪些组成?
- lateral control — steering | 横向控制,如转向
- longitudinal control — braking | 纵向控制,如刹车
- Object and Event Detection and Response (OEDR) | 物体、事件的检测与响应
- planning
- long term (route)
- short term
- miscellaneous
自动驾驶 5 级分类
- L0 - 无自动化
- L1 - 驾驶辅助,如 ACC, LKA
- L2 - 部分自动化,如部分情况下可以执行 Lateral Control / Longitudinal Control
- L3 - 有条件自动化,部分场景下具备 OEDR 能力
- L4 - 高度自动驾驶,具备应变(Fallback)能力
- L5 - 全自动驾驶,unlimited ODD
感知
input -> [Analyze ego motion & env (perceptin)] -> [Decide on and plan a maneuver] -> drive
Object and Event Detection and Response (OEDR)
Goals for perception
- static objects. e.g. road and lane markings, curbs, traffic lights
- dynamic objects (on-road). e.g. vehicles, pedestrians
- ego localization. e.g. position, velocity, orientation
Challenge to perception
- robust detection and segmentation
- sensor uncertainty
- occlusion, reflection
- illumination, lens flare
- weather, precipitation
规划与决策
Planning 分类:
Decision 分类:
- Long term. 从 NY 导航到 LA
- Short term. 现在能否换道到右车道?能否穿过路口?
- Immediate. 加速或刹车?
Reactive Planning
- rule-based
- we have rules that take into account the current state of ego and other objects and give decisions
- e.g. If there is a pedestrian on the road, stop
Predictive Planning
- Make predictions about other vehicles and how they are moving. Then use these predictions to inform our decisions.
- e.g. That car has been stopped for the last 10 seconds. It is going to be stopped for the next few seconds.