LiDAR Simultaneous Localisation and Mapping (LiDAR SLAM)

LiDAR SLAM (Light Detection and Ranging Simultaneous Localisation and Mapping) is a technology that utilizes LiDAR as its primary sensor, enabling mobile robots to determine their real-time position whilst concurrently mapping an unknown environment. This technology serves as one of the key cornerstones for autonomous mobility across intelligent platforms, including autonomous vehicles, automated forklifts, and service robots.

Technical Principles

The fundamental operating principle of LiDAR SLAM is straightforward: the robot's onboard LiDAR emits laser beams and measures the time of flight of reflected signals to calculate precise distance and angular data of surrounding obstacles, generating point cloud data with 3D spatial coordinates. The system subsequently aligns and matches point clouds captured across successive frames to compute relative displacement and attitude variations, thereby completing the robot's self-localisation. Localisation and mapping are interdependent and executed concurrently—precise localisation relies on an existing map, whilst accurate map construction demands reliable pose estimation.

LiDAR units are categorised by their beam channels into single-line (2D) and multi-line (3D) configurations. 2D LiDAR scans obstacles on a single plane, making it ideal for flat-surface applications such as robotic vacuum cleaners; 3D LiDAR captures volumetric 3D point cloud data that reconstructs the precise shape and dimensions of objects, making it well-suited for high-precision demands such as autonomous driving and complex outdoor environments.

Core Algorithm Framework

Modern LiDAR SLAM systems commonly incorporate multi-sensor fusion schemes to boost robustness and precision. Mainstream algorithmic architectures comprise the following core modules:

Front-End Odometry: Estimates pose transformations between successive frames via inter-frame point cloud matching (such as the ICP algorithm and its variants). Research indicates that tightly coupled integration with an IMU (Inertial Measurement Unit) effectively compensates for point cloud motion distortion whilst enhancing real-time processing performance.

Back-End Optimisation: Implements factor graphs or sliding-window optimisation strategies to perform global corrections on cumulative drift generated by the front end. Benchmark algorithms like LOAM (Lidar Odometry and Mapping) and its refined variant, LeGO-LOAM, have become industry-standard solutions in the LiDAR SLAM field.

Loop Closure Detection: Identifies matching features from previously visited areas when the robot re-enters explored territory, establishing spatial constraints to eliminate long-term cumulative drift and ensure global map consistency.

Technical Advantages

Compared to Visual SLAM (V-SLAM), LiDAR SLAM delivers distinctive advantages in the following areas:

• Immune to Ambient Lighting: Operating as an active sensing system, LiDAR performs consistently across low-light environments, total darkness, and direct high-glare sunlight

• High Ranging Precision: With a straightforward measurement error model, LiDAR directly obtains highly accurate geometric distance data, streamlining path planning and navigation

• Broad Operating Versatility: Maintains pinpoint localisation accuracy even in challenging GPS-denied environments, such as tunnels, mining sites, and underground facilities

Development Trends

Current LiDAR SLAM technology is steadily progressing towards deep multi-sensor fusion. A multi-source setup integrating LiDAR, IMU, and vision harnesses LiDAR's high-precision geometric depth, vision's rich textural details, and the IMU's high-frequency motion compensation, delivering optimal adaptability and robustness across intricate operational environments. Furthermore, as mass production continues to drive down the cost of multi-line LiDAR units, 3D LiDAR-based SLAM solutions are seeing accelerated deployment across warehouse logistics, autonomous driving, and related sectors, standing out as a pivotal development direction for the industry.