LiDAR SLAM tutorial

Paper List (TODO)

• 87 PAMI Least-squares fitting of two 3D point sets

  tldr: SVD-based closed form of registration; a basic of basic of the scan matching

• 92 PAMI A method for registration of 3-D shapes

  tldr: a.k.a ICP; a basic of the scan matching

• 97 AR Globally Consistent Range Scan Alignment for Environment Mapping

  tldr: mostly called as “Lu and Milios”; considered as the first work of a scan matching and pose graph optimization-based SLAM.

• 03 IROS The Normal Distributions Transform: A New Approach to Laser Scan Matching

  tldr: NDT registration

• 06 IJRR Square Root SAM: Simultaneous Localization and Mapping via Square Root Information Smoothing

  tldr: from probabilistic nature to least square formulation of SLAM for smoothing (i.e., modification of past poses)

• 07 JFR Scan registration for autonomous mining vehicles using 3D-NDT

  tldr: 3D version of NDT registration

• 08 TRO iSAM: Incremental Smoothing and Mapping

  tldr: incremental SAM and an open source library

• 09 ICRA Real-Time Correlative Scan Matching

  tldr: prof. Olson; later affects to Cartographer, etc.

• 09 ICRA Fast Point Feature Histograms (FPFH) for 3D Registration

  tldr: FPFH (the most famous 3D local descriptor) registration

• 09 RSS Generalized-ICP

  tldr: uncertainty-embedded ICP (probabilistic perspective)

• 10 ITSM A Tutorial on Graph-Based SLAM

  tldr: Grisetti’s must-read tutorial

• 11 IV Velodyne SLAM

  tldr: an early work of modern 3D scanning LiDAR-based motion estimation

• 12 TRO Zebedee: Design of a Spring-Mounted 3-D Range Sensor with Application to Mobile Mapping

  tldr: mobile mapping system and IMU fusion

• 12 RAM Tutorial: Point Cloud Library: Three-Dimensional Object Recognition and 6 DOF Pose Estimation

  tldr: PCL tutorial, but not much delved into the SLAM perspective.

• 12 IJRR iSAM2: Incremental smoothing and mapping using the Bayes tree

  tldr: in GTSAM 4.0, iSAM2 (not iSAM1) is currently a de-facto default factor graph optimizer.

• 13 ICRA Robust Odometry Estimation for RGB-D Cameras

  tldr: a.k.a DVO; this is not an actually LiDAR thing, but to understand the effectiveness of direct alignment rather ICP

• 13 IROS Dense Visual SLAM for RGB-D Cameras

  tldr: a SLAM version (i.e., including loop closures) of the DVO; studying RGB-D SLAMs is also worthy for LiDAR guys because they frequently considers the both a photometric error and a geometric error.

• 13 AR Challenging data sets for point cloud registration algorithms

  tldr: a.k.a the open library: Libpointmatcher

• 14 RSS LOAM: Lidar Odometry and Mapping in Real-time

  tldr: THE LOAM; (surface and corner) feature matching for frame-to-frame registration and frame-to-map refinement

• 15 ICRA Visual-lidar Odometry and Mapping: Low-drift, Robust, and Fast

  tldr: Visual + LOAM

• 15 ICRA Initialization Techniques for 3D SLAM: a Survey on Rotation Estimation and its Use in Pose Graph Optimization

  tldr: LiDAR SLAM usually have more opportunity to think of the better pose graph optimization because it directly measures the depth (rather easier front-end than visual domain).

• 15 RAM Registration with the Point Cloud Library: A Modular Framework for Aligning in 3-D

  tldr: PCL tutorial for registration

• 15 IROS NICP: Dense normal based point cloud registration

  tldr: as already in the title, dense normal

• 16 ICRA Real-time loop closure in 2D LIDAR SLAM

  tldr: the Google Cartographer’s paper

• 16 SSRR ICP-based pose-graph SLAM

  tldr: an almost standard framework of scan matching- and pose-graph-based LiDAR SLAM

• 16 IROS M2DP: A novel 3D point cloud descriptor and its application in loop closure detection

  tldr: place descriptor using a single LiDAR scan

• 16 BookChapter World modeling

  tldr: various representations of a surrounding environment; selecting a proper representation of the environment is important and it determines the state estimation ways.

• 18 RSS Efficient Surfel-Based SLAM using 3D Laser Range Data in Urban Environments

  tldr: a.k.a SuMa; projective view rendering, and open source

• 18 ICRA IMLS-SLAM: scan-to-model matching based on 3D data

  tldr: sophisticated feature selections, not real-time but for the accuracy, this is considered as a SOTA

• 18 ICRA Elastic LiDAR Fusion: Dense Map-Centric Continuous-Time SLAM

  tldr: (continuous-time) non-rigid map deformation (see 15 RSS ElasticFusion also)

• 18 ICRA Efficient Continuous-time SLAM for 3D Lidar-based Online Mapping

  tldr: a hierarchical continuous-time LiDAR SLAM

• 18 IROS LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain

  tldr: range image-based fast feature selection for LOAM; and open source

• 18 IROS LIPS: LiDAR-Inertial 3D Plane SLAM

  tldr: leveraging plane for LINS system

• 18 IROS Scan Context: Egocentric Spatial Descriptor for Place Recognition Within 3D Point Cloud Map

  tldr: A visibility-based place descriptor for fast and robust place recognition, and open source

• 19 A-LOAM (code only)

  tldr: a well-implemented LOAM algorithm (the original LOAM author closed the official code), and open source

• 19 ICRA Tightly Coupled 3D Lidar Inertial Odometry and Mapping

  tldr: a.k.a lio-mapping; imu tight fusion but practically slow, and open source

• 19 IV DeLiO: Decoupled LiDAR Odometry

  tldr: rotation and translation are decoupled

• 19 IJRR SegMap: Segment-based mapping and localization using data-driven descriptors

  tldr: deep segment feature learning for LiDAR place recognition

• 19 IROS SuMa++: Efficient LiDAR-based Semantic SLAM

  tldr: merging semantic information into SuMa

• 20 AR DVL-SLAM: sparse depth enhanced direct visual-LiDAR SLAM

  tldr: enhanced visual SLAM by LiDAR data

• 20 RSS OverlapNet: Loop Closing for LiDAR-based SLAM

  tldr: learning two scan’s overlap and integrated it into the modern probabilistic SLAM system.

• 20 IROS LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping

  tldr: IMU fusion (tightly) of LOAM, and open source.

• 20 IROS SpoxelNet: Spherical Voxel-based Deep Place Recognition for 3D Point Clouds of Crowded Indoor Spaces

  tldr: Deep LiDAR feature learning for place recognition and robust to occlusions

• 20 IROS Semantic Graph Based Place Recognition for 3D Point Clouds

  tldr: Summarizing a place with a single semantic graph. The matching part is also deep (SegMap didn’t).

• 20 IROS A Fast and Robust Place Recognition Approach for Stereo Visual Odometry Using LiDAR Descriptors

  tldr: LiDAR descriptors are also good for stereo-camera-based place recognition