Hardware
1. LiDAR
Light Detection And Ranging
Yields point cloud
Parameters:
- FOV (Field of View) (horizontal, vertical, ...)
- Angle resolution
- Spinning rate: e.g. 5-20Hz
Applications:
- Mapping?
- Perception
- Localization
Different types:
- Mechanical rotation (mostly used; easy to be broken, not reliable)
- MEMS-mirror: just rotate a small mirror (reliable)
- Optical phase array: doesn't move; use electricity to control the direction of laser; more challenging production (but more reliable)
- Flash LiDAR
Raw data:
- distance
- vertical or horizontal angle
- timestamp
Evolution
- Smaller & Cheaper
- Denser data
- Solid state
Advantages:
- Accurate depth (more accurate than radar)
- Active sensor
- Robust against ambient light
Disadvantages:
- Sparse (1080 of camera vs 64)
- Bulky (taking up much space; large and unwieldy)
2. Camera
Applications:
- Obstacle detection / classification
- Traffic light (sometimes not easy)
Important specs:
- Dynamic range: cover different levels of brightness
- Global shutter vs rolling shutter (cheaper; can read the data line by line; need to be corrected)
Advantages:
- Low cost
- Rich data
- Color
Disadvantages:
- Only 2D info (need to reconstruct 3D scenes)
- Affected by ambient light
- Computation burden
- Storage burden
- Bandwidth burden
3. Radar
Radio detection and ranging
Determine the range, angle and velocity of objects
Emits chirps?
short range mode or far range mode
Advantages:
- Usable under bad weather
- Good longitude speed accuracy
Disadvantages:
- Bad lateral speed accuracy
- False positives
4. GPS
Real Time Differential GPS
- DGPS receiver
- DGPS site
5. IMU
Inertial Measurement Unit
measures relative ego-position and ego-pose
drift over time
6. Compute System
SoC: System on a chip
Last update:
October 17, 2021
Authors: