LiDAR

Practical

LiDAR (Light Detection and Ranging) uses laser pulses to measure distances with millimeter precision. By scanning in 2D or 3D, it creates detailed point clouds of the environment, essential for autonomous navigation, mapping, and obstacle detection.

How LiDAR Works

1. Emit laser pulse ──────────► Object
2. Pulse reflects ◄──────────── Object
3. Measure time-of-flight
4. Calculate distance: d = (c × t) / 2

c = speed of light
t = round-trip time

Modern LiDARs emit thousands of pulses per second, building 3D maps in real-time.

LiDAR Types

Mechanical Spinning

A rotating mirror or assembly scans the laser across the scene.

Pros	Cons
360° field of view	Moving parts wear out
Mature technology	Expensive ($4,000-$75,000)
High accuracy	Bulky form factor

Examples: Velodyne VLP-16 (now sold through Ouster), Ouster OS1

Solid-State / Hybrid

No external moving parts. Uses phased arrays, MEMS mirrors, or internal rotating mirrors.

Pros	Cons
Compact, rugged	Limited field of view
Lower cost	Newer technology
No mechanical wear	May need multiple units for 360°

Examples: Livox Mid-360 (hybrid), Hesai AT128

Flash LiDAR

Illuminates entire scene at once, like a camera flash.

Pros	Cons
No scanning needed	Shorter range
Very fast	Lower resolution
Simple design	Higher power consumption

Examples: Intel RealSense L515 (discontinued Feb 2022; consider D455/D435i stereo depth cameras as alternatives)

Key Specifications

Spec	Description	Typical Range
Range	Maximum detection distance	10-300m
Points/sec	Measurement rate	300K-2M
Channels	Vertical resolution (spinning)	16-128
FoV	Field of view	90°-360°
Angular resolution	Angle between points	0.1°-0.4°
Accuracy	Distance error	±2-5cm

Robotics Applications

SLAM and Mapping

LiDAR provides accurate distance measurements for building maps:

• 2D LiDAR → 2D occupancy grids (mobile bases)
• 3D LiDAR → 3D point cloud maps (autonomous vehicles)

Obstacle Detection

Reliable detection regardless of lighting:

• Works in complete darkness
• Not affected by shadows or glare
• Precise distance to obstacles

Localization

Match current scan to known map:

• Iterative Closest Point (ICP)
• Normal Distributions Transform (NDT)

LiDAR vs Cameras vs Radar

Aspect	LiDAR	Camera	Radar
Range accuracy	Excellent	Poor (needs depth estimation)	Good
Works in dark	Yes	No	Yes
Works in fog/rain	Degraded	Degraded	Good
Color/texture	No	Yes	No
Cost	High	Low	Medium
Point density	High	Very high (pixels)	Low

Most autonomous systems use sensor fusion — combining all three.

ROS 2 Integration

Standard message type: sensor_msgs/PointCloud2

# View LiDAR data
ros2 topic echo /scan            # 2D LaserScan
ros2 topic echo /points          # 3D PointCloud2

# Visualize in RViz2
ros2 run rviz2 rviz2
# Add PointCloud2 display, set topic to /points

Common Packages

Package	Purpose
pointcloud_to_laserscan	Convert 3D to 2D
laser_filters	Filter noisy points
pcl_ros	Point Cloud Library integration
nvblox	GPU-accelerated 3D reconstruction

NVIDIA Isaac Integration

Isaac ROS 4.0 (current release, requires ROS 2 Jazzy) provides GPU-accelerated LiDAR processing:

# nvblox: Real-time 3D reconstruction from depth cameras and/or 3D LiDAR
# Isaac ROS packages:
# - isaac_ros_nvblox
# - isaac_ros_pointcloud_utils

Isaac Sim can simulate LiDAR sensors with realistic noise models.

Popular Sensors by Application

Mobile Robots (Indoor)

• RPLidar A1/A2 — $100-200, 2D, 360°
• Livox Mid-360 — $1,000, 3D solid-state

Mobile Robots (Outdoor)

• Ouster OS0/OS1 — $4,000-8,000, 3D, 360°
• Velodyne VLP-16 — $4,000, 3D, 16-channel (via Ouster)

Autonomous Vehicles

• Hesai AT128 — $1,000+, automotive-grade
• Hesai FT120 — Solid-state, blind-spot detection

Drones

• Livox Mid-40 — Lightweight, compact
• Velodyne VLP-16 — If payload allows

Related Terms

• SLAM — Uses LiDAR for mapping
• Cameras — Complementary sensor
• Isaac ROS — GPU-accelerated processing

Sources

• Ouster VLP-16 — VLP-16 specs and availability
• Ouster OS1 — OS1 specifications
• Livox Mid-360 — Mid-360 specs and hybrid design
• Hesai AT128 — AT128 automotive LiDAR specs
• SLAMTEC RPLidar — RPLidar A1/A2 specifications
• NVIDIA Isaac ROS — Isaac ROS packages and version info
• ROS 2 Releases — Current ROS 2 distributions