gaoxiang12/faster-lio

Faster-LIO: Lightweight Tightly Coupled Lidar-inertial Odometry using Parallel Sparse Incremental Voxels

What it solves

FasterLIO is a lightweight Lidar-inertial odometry (LIO) system designed for high-speed lidar pose tracking and point cloud mapping. It addresses the computational bottlenecks of existing LIO systems to provide significantly faster processing speeds, enabling real-time performance on a wider range of hardware.

How it works

Built upon the foundation of FastLIO2, FasterLIO implements a tightly coupled Lidar-inertial odometry approach. It utilizes parallel sparse incremental voxels (iVox) to optimize the mapping and tracking process. The system supports different iVox node types, such as linear iVox and PHC iVox, which can be specified during compilation to tailor performance.

Who it’s for

This project is intended for robotics researchers and developers working with lidar-equipped autonomous systems that require high-frequency pose estimation and mapping, specifically those using solid-state lidars or typical 32-line spinning lidars.

Highlights

  • High Performance: Achieves 1.5-2x speed increase over FastLIO2, reaching nearly 1k-2k Hz for solid-state lidars and over 100 Hz for spinning lidars.
  • Flexible Deployment: Supports both offline processing of bag files and online real-time execution via ROS.
  • Broad Hardware Support: Tested on various CPU architectures, including AMD and Intel processors.
  • Detailed Diagnostics: Provides comprehensive run-time breakdown of time usage for components like IEKF solve, lidar matching, and incremental mapping.

Related

  • Project
    pypose/pypose

    A PyTorch-based library that combines deep perceptual models with physics-based optimization for robot learning, providing tools for Lie groups, filtering, and second-order optimization.

  • Project
    ARISE-Initiative/robomimic

    A framework for robot learning from demonstration that provides standardized datasets and offline learning algorithms to make robot manipulation research accessible and reproducible.

  • Project
    symforce-org/symforce

    A symbolic computation and code generation library for robotics that transforms mathematical expressions into optimized C++ code for tasks like SLAM and motion planning.

  • Project
    ShisatoYano/AutonomousVehicleControlBeginnersGuide

    A collection of Python sample codes and simulations covering localization, mapping, path planning, and path tracking for autonomous vehicle control.