LiDAR SLAM 논문 리스트

시작하기 전…

이 글의 출처는 Giseop Kim 님의 Notion 글입니다. 그냥 다 그대로 가져온 수준이에요(…)

기섭님 블로그, 유투브에도 좋은 슬램 관련 글과 영상이 있습니다.

Paper List

• 1987 PAMI Least-squares fitting of two 3D point sets
  • SVD-based closed form of registration; a basic of basic of the scan matching
• 92 PAMI A method for registration of 3-D shapes
  • ICP - a basic of the scan matching
• 97 AR Globally Consistent Range Scan Alignment for Environment Mapping
  • 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
  • NDT registration
• 06 IJRR Square Root SAM: Simultaneous Localization and Mapping via Square Root Information Smoothing
  • 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
  • 3D version of NDT registration
• 08 TRO iSAM: Incremental Smoothing and Mapping
  • incremental SAM and an open source library
• 09 ICRA Real-Time Correlative Scan Matching
  • prof. Olson; later affects to Cartographer, etc
• 09 ICRA Fast Point Feature Histograms (FPFH) for 3D Registration
  • FPFH (the most famous 3D local descriptor) registration
• 09 RSS Generalized-ICP
  • uncertainty-embedded ICP (probabilistic perspective)
• 10 ITSM A Tutorial on Graph-Based SLAM
  • Grisetti’s must-read tutorial
• 11 IV Velodyne SLAM
  • 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
  • mobile mapping system and IMU fusion
• 12 RAM Tutorial: Point Cloud Library: Three-Dimensional Object Recognition and 6 DOF Pose Estimation
  • PCL tutorial, but not much delved into the SLAM perspective.
• 12 IJRR iSAM2: Incremental smoothing and mapping using the Bayes tree
  • 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
  • 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
  • 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
  • a.k.a the open library: Libpointmatcher*
• 14 RSS LOAM: Lidar Odometry and Mapping in Real-time
  • 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
  • Visual + LOAM
• 15 ICRA Initialization Techniques for 3D SLAM: a Survey on Rotation Estimation and its Use in Pose Graph Optimization
  • 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
  • PCL tutorial for registration
• 15 IROS [NICP: Dense normal based point cloud registration
  • as already in the title, dense normal](http://jacoposerafin.com/wp-content/uploads/serafin15iros.pdf)
• 16 ICRA Real-time loop closure in 2D LIDAR SLAM
  • the Google Cartographer’s paper
• 16 SSRR ICP-based pose-graph SLAM
  • 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
  • place descriptor using a single LiDAR scan
• 16 BookChapter World modeling
  • 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
  • a.k.a SuMa; projective view rendering, and open source
• 18 ICRA IMLS-SLAM: scan-to-model matching based on 3D data
  • 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
  • (continuous-time) non-rigid map deformation (see 15 RSS ElasticFusion also)
• 18 ICRA Efficient Continuous-time SLAM for 3D Lidar-based Online Mapping
  • a hierarchical continuous-time LiDAR SLAM
• 18 IROS LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain
  • range image-based fast feature selection for LOAM; and open source
• 18 IROS LIPS: LiDAR-Inertial 3D Plane SLAM
  • leveraging plane for LINS system
• 18 IROS Scan Context: Egocentric Spatial Descriptor for Place Recognition Within 3D Point Cloud Map
  • A visibility-based place descriptor for fast and robust place recognition, and open source
• 19 A-LOAM (code only)
  • 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
  • a.k.a lio-mapping; imu tight fusion but practically slow, and open source
• 19 IV DeLiO: Decoupled LiDAR Odometry
  • rotation and translation are decoupled
• 19 IJRR SegMap: Segment-based mapping and localization using data-driven descriptors
  • deep segment feature learning for LiDAR place recognition
• 19 IROS SuMa++: Efficient LiDAR-based Semantic SLAM
  • merging semantic information into SuMa
• 20 AR DVL-SLAM: sparse depth enhanced direct visual-LiDAR SLAM
  • enhanced visual SLAM by LiDAR data
• 20 RSS OverlapNet: Loop Closing for LiDAR-based SLAM
  • 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
  • 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
  • Deep LiDAR feature learning for place recognition and robust to occlusions
• 20 IROS Semantic Graph Based Place Recognition for 3D Point Clouds
  • 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
  • LiDAR descriptors are also good for stereo-camera-based place recognition