LiDAR Overview

What Is LiDAR

LiDAR (Light Detection and Ranging) is an active sensing technology that measures distances using laser pulses. It constructs precise 3D models of the environment by emitting lasers and receiving reflected signals, making it one of the most important sensors in robot perception systems.

Ranging Principles

Time-of-Flight (ToF)

The most fundamental LiDAR ranging principle. A short laser pulse is emitted, and its round-trip time is measured:

\[ d = \frac{c \cdot t}{2} \]

Where:

• \(d\) is the target distance
• \(c\) is the speed of light (\(\approx 3 \times 10^8 \, \text{m/s}\))
• \(t\) is the round-trip time of the laser pulse

Characteristics:

• Simple principle, mature implementation
• Large ranging distance (up to hundreds of meters)
• Accuracy limited by time measurement resolution (\(\Delta d = \frac{c \cdot \Delta t}{2}\))
• Widely used in mechanical LiDARs (e.g., Velodyne)

Phase-Shift Method

Emits a continuously modulated laser and calculates distance by measuring the phase difference between emitted and reflected signals:

\[ d = \frac{c \cdot \Delta\varphi}{4\pi f} \]

Where:

• \(\Delta\varphi\) is the phase difference
• \(f\) is the modulation frequency

Characteristics:

• Higher accuracy (millimeter-level)
• Relatively shorter ranging distance
• Needs to resolve phase ambiguity (multi-frequency modulation)
• Commonly used in industrial measurement-grade LiDARs

Frequency-Modulated Continuous Wave (FMCW)

Emits a continuous laser whose frequency varies linearly with time. Distance is determined by the beat frequency between the reflected signal and a local reference signal:

\[ d = \frac{c \cdot f_{\text{beat}}}{2B / T} \]

Where:

• \(f_{\text{beat}}\) is the beat frequency
• \(B\) is the frequency sweep bandwidth
• \(T\) is the sweep period

Characteristics:

• Can simultaneously obtain distance and velocity information (Doppler effect)
• Strong resistance to ambient light interference
• High ranging accuracy
• High technical complexity and cost
• Representatives: Aeva, SiLC Technologies

LiDAR Type Classification

By Scanning Method

graph TD
    A[LiDAR Types] --> B[Mechanical Rotating]
    A --> C[Semi-Solid-State]
    A --> D[Pure Solid-State]

    B --> B1[Single/Multi-line Rotation<br>Velodyne VLP-16<br>Ouster OS1]

    C --> C1[MEMS Micro-Mirror<br>RoboSense]
    C --> C2[Prism Rotation<br>Livox Mid-360]

    D --> D1[OPA Optical Phased Array]
    D --> D2[Flash LiDAR]
    D --> D3[FMCW Solid-State]

Mechanical Rotating LiDAR

Feature	Description
Principle	Laser emitter/receiver module rotates with a motor
FOV	Horizontal 360 degrees, vertical depends on number of lines
Advantages	Omnidirectional scanning, mature technology
Disadvantages	Mechanical wear, large size, high cost
Representative	Velodyne VLP-16/32/64/128

Semi-Solid-State LiDAR

Feature	Description
Principle	Small mirror or prism scanning
FOV	Limited angle (typically 60-120 degrees)
Advantages	Small size, higher reliability
Disadvantages	Limited FOV
Representative	Livox Mid-360, HAP

Pure Solid-State LiDAR

Feature	Description
Principle	No moving parts whatsoever
FOV	Limited angle
Advantages	High reliability, low cost potential, mass-producible
Disadvantages	Technology still maturing
Representative	Cepton, Ibeo

Key Performance Metrics

Metric	Description	Typical Range
Range	Maximum detectable distance	12m (low-end) - 300m+ (high-end)
Range Accuracy	Distance measurement error	+/-1cm - +/-3cm
Angular Resolution	Angle between adjacent scan lines/points	0.1 - 2 degrees
Field of View (FOV)	Horizontal and vertical scan range	H: 60-360 deg, V: 15-90 deg
Scan Frequency	Scans completed per second	5Hz - 20Hz
Point Rate	Points generated per second	10K - 2.4M points/s
Number of Returns	Returns per pulse	1 - 5
Wavelength	Laser wavelength	905nm (near-IR) or 1550nm (eye-safe)
Power Consumption	Operating power	5W - 30W
Protection Rating	Environmental protection	IP65 - IP69K

LiDAR Selection for Different Robot Platforms

Indoor Mobile Robots / Robot Vacuums

• Recommended: 2D LiDAR (RPLIDAR A1/A2, YDLIDAR X4)
• Reason: Low cost, low power; 2D SLAM sufficient for navigation
• Typical setup: Single-line 360-degree LiDAR mounted on top of the robot

Service Robots / AGVs

• Recommended: 2D LiDAR (safety-rated) + optional 3D LiDAR
• Reason: Safety certification needed (e.g., SICK TiM series); obstacle avoidance requirements
• Typical setup: One safety LiDAR front and rear + top navigation LiDAR

Autonomous Driving Vehicles

• Recommended: Multi-line 3D LiDAR or solid-state LiDAR combinations
• Reason: High-precision 3D environment perception, long-range detection needed
• Typical setup: Main LiDAR on roof + corner blind-spot LiDARs

Drones

• Recommended: Lightweight solid-state LiDAR (e.g., Livox Mid-360)
• Reason: Weight and power constraints
• Typical setup: Downward or forward-facing mount

Quadruped Robots

• Recommended: Lightweight 3D LiDAR (Livox Mid-360, Ouster OS0)
• Reason: Dynamic environment perception, terrain mapping
• Typical setup: Head or back-mounted

LiDAR Data Processing Pipeline

graph LR
    A[Raw Data Acquisition] --> B[Data Preprocessing]
    B --> C[Feature Extraction]
    C --> D[Advanced Applications]

    B --> B1[Noise Filtering]
    B --> B2[Motion Compensation]
    B --> B3[Coordinate Transform]

    C --> C1[Plane Detection]
    C --> C2[Cluster Segmentation]
    C --> C3[Edge/Corner Extraction]

    D --> D1[SLAM]
    D --> D2[Object Detection]
    D --> D3[Map Building]

Laser Safety Classes

Class	Description	LiDAR Application
Class 1	Safe under all operating conditions	Most consumer/industrial LiDARs
Class 1M	Safe for naked eye, potentially unsafe with optical instruments	Some long-range LiDARs
Class 3R	Low risk of direct viewing	Few high-power measurement LiDARs

Eye Safety

905nm wavelength LiDARs must strictly control power to ensure eye safety. 1550nm wavelength is safer for human eyes (absorbed by the cornea rather than reaching the retina), but detector costs are higher.

Comparison with Other Sensors

Feature	LiDAR	Camera	mmWave Radar	Ultrasonic
Accuracy	High (cm-level)	Medium	Medium	Low
Range	Far (~300m)	Far	Far (~250m)	Near (~5m)
3D Information	Native 3D	Requires algorithms	Limited	No
Affected by Lighting	Slightly	Severely	No	No
Affected by Weather	Rain/fog significant	Rain/fog significant	Better	Better
Cost	High	Low	Medium	Very low
Texture/Color	No	Yes	No	No
Power	Medium	Low	Low	Very low

References

• Velodyne LiDAR Technical White Papers
• Livox Technical Documentation
• Introduction to Autonomous Mobile Robots - Siegwart et al.
• ROS2 LiDAR Driver Package Documentation

---

评论