[자율주행 스터디] Introduction to Self-Driving cars - Week 1-2

Week 1

Course info.

• https://www.coursera.org/learn/intro-self-driving-cars?
  • UNIVERSITY OF TORONTO

---

Lesson 2. Requirements for Perception

Contents

• What is perception?
• Goals for perception
  • Static, dynamic objects
  • Ego requirements
• Challenges to perception

Roughly...

Input -> Perception (Analyze ego motion & environment)

-> Planning (Decide on and plan a maneuver) -> Drive

---

What is perception?

• We want to make sense of the environment and ourselves
• Two things
  • Identification
  • Understanding motion
• Why?
  • To inform our driving decisions

---

Goals for perception

• Objects
  • On-road
    • Static objects
      • Road and lane markings (도로에 그려진 표시들)
      • Construction signs (공사 관련 신호들)
      • Obstructions (시야를 가리는 장애물들)
    • Dynamic objects
      • Vehicles
        • 4 wheelers (cars, trucks)
        • 2 wheelers (motorbikes, bicycles)
      • Pedestrians
  • Off-road
    • Static objects
      • Curbs (연석)
      • Traffic lights (신호등)
      • Road signs (속도 제한 표지판, 주정차 금지 표지판 같은 신호들)
• Ego localization
  • Position
  • Velocity, acceleration
  • Orientation, angular motion

---

Challenges to perception

• Robust detection and segmentation
  • Sensor uncertainty
    • Occlusion, reflection
    • Illumination, lens flare
    • Weather precipitation

원래 표시는 위와 같지 않지만, 계층을 굳이 나눠보면 위와 같이 나뉠 것 같다.

강건한 감지 및 분류를 위해서는 결국 입력단(센서)에서 제공되는 정보의 정확성이 중요하다. 센서가 정확한 정보를 주지 못하는 상황들을 아래 계층으로 매칭했다.