BIEVR-LIO-SLAM - LiDAR-Inertial SLAM & Localization

Overview

BIEVR-LIO-SLAM turns the BIEVR-LIO LiDAR–Inertial Odometry framework into a complete SLAM and localization system.

The base odometry uses a high-resolution, voxel-wise oriented height image map to exploit subtle geometric variations in challenging, degenerate environments. On top of it, this project adds loop closure, pose-graph optimization, persistent mapping and global relocalization.

Architecture

The guiding design principle is that all SLAM modules live downstream of the odometry, connected through a single observer hook.

• The observer receives only (timestamp, body pose, undistorted point cloud) per frame
• Corrections are published alongside the odometry and never fed back into it
• The odometry therefore behaves identically whether the SLAM modules run or not

Because the coupling is that thin, the same stack ports to any LOAM-like LIO system (FAST-LIO, LIO-SAM, …) that can expose those three outputs.

Method

• Place recognition — Scan Context descriptors (rotation-invariant polar ring/sector image)
• Back-end — GTSAM pose-graph optimization with incremental iSAM2
• Keyframing — gating by translation and rotation thresholds
• Robustness — Cauchy noise model on loop constraints to reject false positives
• Localization — ICP tracking against a prior map in the map frame, with automatic Scan Context relocalization or manual seeding from RViz 2D Pose Estimate
• Scalability — tile-based map caching, so maps larger than available RAM stay usable

Key Features

• Loop closure with pose-graph optimization to correct trajectory drift
• Dual-mode operation: mapping (build & save) and localization (track against a prior map)
• Saved map bundles (cloud.pcd, scan_context.bin, poses_tum.txt, meta.yaml)
• Works with bare .pcd maps produced by other SLAM systems

Implementation

• Language — C++ (Eigen, Ceres, PCL, GTSAM 4.2, yaml-cpp)
• Middleware — ROS 2 Jazzy (full SLAM support); ROS 1 Noetic (odometry only)
• Sensors — standard industrial LiDARs plus Livox gen1/gen2
• Datasets — ready-made sensor configs for ENWIDE, Newer College, GEODE, MARS-LVIG and GrandTour

Use Cases

• Large-scale mapping and map reuse
• Long-term localization in GPS-denied environments
• Field and industrial robotics, autonomous navigation

🔗 Project Access

👉 View on GitHub 👉 Watch the Demo