# Geometric Multi-Session Map Merging with Learned Local Descriptors

arXiv: https://arxiv.org/abs/2512.24384 · HTML: https://mlsanity.com/paper/2512.24384/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 38 references that are in the map.
- First posted on arXiv 30 Dec 2025.

## Builds on (most cited first, 22 of 22)

- [KPConv: Flexible and Deformable Convolution for Point Clouds](https://mlsanity.com/paper/1904.08889/) (2019)
- [Circle Loss: A Unified Perspective of Pair Similarity Optimization](https://mlsanity.com/paper/2002.10857/) (2020)
- [Geometric Transformer for Fast and Robust Point Cloud Registration](https://mlsanity.com/paper/2202.06688/) (2022)
- [PointNetVLAD: Deep Point Cloud Based Retrieval for Large-Scale Place Recognition](https://mlsanity.com/paper/1804.03492/) (2018)
- [LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping](https://mlsanity.com/paper/2007.00258/) (2020)
- [FAST-LIO2: Fast Direct LiDAR-inertial Odometry](https://mlsanity.com/paper/2107.06829/) (2021)
- [MinkLoc3D: Point Cloud Based Large-Scale Place Recognition](https://mlsanity.com/paper/2011.04530/) (2020)
- [OverlapTransformer: An Efficient and Rotation-Invariant Transformer Network for LiDAR-Based Place Recognition](https://mlsanity.com/paper/2203.03397/) (2022)
- [LCDNet: Deep Loop Closure Detection and Point Cloud Registration for LiDAR SLAM](https://mlsanity.com/paper/2103.05056/) (2021)
- [Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems](https://mlsanity.com/paper/2106.14386/) (2021)
- [LoGG3D-Net: Locally Guided Global Descriptor Learning for 3D Place Recognition](https://mlsanity.com/paper/2109.08336/) (2021)
- [SegMap: Segment-based mapping and localization using data-driven descriptors](https://mlsanity.com/paper/1909.12837/) (2019)
- [Direct LiDAR-Inertial Odometry: Lightweight LIO with Continuous-Time Motion Correction](https://mlsanity.com/paper/2203.03749/) (2022)
- [LAMP: Large-Scale Autonomous Mapping and Positioning for Exploration of Perceptually-Degraded Subterranean Environments](https://mlsanity.com/paper/2003.01744/) (2020)
- [LAMP 2.0: A Robust Multi-Robot SLAM System for Operation in Challenging Large-Scale Underground Environments](https://mlsanity.com/paper/2205.13135/) (2022)
- [RSKDD-Net: Random Sample-based Keypoint Detector and Descriptor](https://mlsanity.com/paper/2010.12394/) (2020)
- [AutoMerge: A Framework for Map Assembling and Smoothing in City-scale Environments](https://mlsanity.com/paper/2207.06965/) (2022)
- [Globally Consistent 3D LiDAR Mapping with GPU-accelerated GICP Matching Cost Factors](https://mlsanity.com/paper/2109.07073/) (2021)
- [FRAME: Fast and Robust Autonomous 3D point cloud Map-merging for Egocentric multi-robot exploration](https://mlsanity.com/paper/2301.09213/) (2023)
- [Simultaneous merging multiple grid maps using the robust motion averaging](https://mlsanity.com/paper/1706.04463/) (2017)
- [DCL-SLAM: A Distributed Collaborative LiDAR SLAM Framework for a Robotic Swarm](https://mlsanity.com/paper/2210.11978/) (2022)
- [Direct LiDAR-Inertial Odometry and Mapping: Perceptive and Connective SLAM](https://mlsanity.com/paper/2305.01843/) (2023)

## Abstract (arXiv)

Multi-session map merging is crucial for extended autonomous operations in large-scale environments. In this paper, we present GMLD, a learning-based local descriptor framework for large-scale multi-session point cloud map merging that systematically aligns maps collected across different sessions with overlapping regions. The proposed framework employs a keypoint-aware encoder and a plane-based geometric transformer to extract discriminative features for loop closure detection and relative pose estimation. To further improve global consistency, we include inter-session scan matching cost factors in the factor-graph optimization stage. We evaluate our framework on the public datasets, as well as self-collected data from diverse environments. The results show accurate and robust map merging with low error, and the learned features deliver strong performance in both loop closure detection and relative pose estimation.