Operating in previously visited environments is becoming increasingly crucial for autonomous systems, with direct applications in autonomous driving, surveying, and warehouse or household robotics. This repeated exposure to observing the same areas poses significant challenges for mapping and localization across sessions, particularly in deciding when a prior model is sufficient for reliable localization and when new mapping is required. In this work, we propose a novel multi-session framework that builds on map-based localization, in contrast to the common practice of greedily running full SLAM sessions and trying to find correspondences between the resulting maps. The core contribution is a principled, topology-driven mechanism to detect multi-session mapping needs from the pose-graph structure. Specifically, our approach uses spectral connectivity metrics on the joint pose-graph to identify disconnections and weakly constrained regions, and selectively triggers mapping and loop closing only when the pose-graph topology indicates insufficient structural support. The resulting map and pose-graph are seamlessly integrated into the existing model, reducing accumulated error and enhancing global consistency while avoiding redundant remapping. We validate our method on overlapping sequences from datasets and demonstrate its effectiveness in a real-world mine-like environment.
Figures & tables
Fig. 1 : Map and pose-graph built after processing three overlapping sessions of the Newer College Extension dataset [ 6 ] ( Quad M , Quad H , Cloister ); each depicted in a different color for visualization.
Fig. 2 : Overview of the proposed method.
Fig. 3 : Visualization of the graphs used in the multi-session framework. The active graph (blue) is built from selected key frames of the odometry graph (gray) and their condensed measurements, together with vertices from the reference graph (orange) connected via inter-session correspondences. When triggered by the topological decision-making module ( e.g. , in unmapped regions), a subset of the active graph (candidate graph, green) is integrated into the reference model. LC and global PGO are performed to ensure consistency.
Fig. 4 : Illustration of the evolution over time of the weighted connectivity indices (blue: dˉ , red: λˉ2 ) during the traversal of previously unmapped regions. Stage I corresponds to traversal within previously mapped regions, where localization is performed. Stage II indicates entry into unmapped regions, resulting in disconnection between the reference and active graphs, and necessitating mapping. Stage III marks reconnection with the prior model and re-establishes map-based localization. The topmost images illustrate examples of these phases: the orange map and PG are the reference model, while the blue ones correspond to the online point cloud and sliding graph.
Fig. 5 : Multi-session mapping results in NC and NCE datasets: Quad-e (orange), Quad-m (teal), Quad-h (red), Cloister (purple), and Short (blue). Notably, overlapping regions are modeled only once (both in the map and the PG).
Method / Sequence
Quad-e
Quad-m
Quad-h
Cloister
Short
Avg.
SLAM
LIO-SAM [ 21 ]
0.074
0.067
0.13
0.074
0.41
0.15
NV-LIOM [ 22 ]
0.076
0.076
0.18
0.077
0.48
0.18
LG-SLAM [ 19 ]
0.066
0.072
0.090
0.076
0.38
0.14
Map-based Localization
HDL-Loc [ 23 ]
–
0.10
×
×
×
×
G-LOC [ 13 ]
–
0.067
0.087
0.21
1.61
0.39
Ours
0.066
0.067
0.082
0.077
0.43
0.14
TABLE I: ATE RMSE (m) in NC and NCE dataset. SLAM methods build separate maps for each sequence. Map-based localization methods employ Quad-e as reference map. Our method incrementally concatenates sessions (left to right). Best results are bold and second best are underlined. × indicates an unrecoverable failure.
Fig. 6 : Evolution of the weighted average node degree (blue) and Fiedler value (red) over time while processing Quad-h sequence.
Fig. 7 : Visualization of the joint PG after processing Quad-e and Quad-h . Orange nodes belong to the prior graph ( Quad-e ), while blue ones are the new nodes incorporated to the model during Quad-h sequence. Red edges represent the weakest part of the graph.
Method / Seq.
Individual
Combination
S1
S2
S1 (given S2 )
S2 (given S1 )
M2F (from [ 3 ] )
0.28
0.46
0.27
0.49
F2F (from [ 3 ] )
0.28
0.46
0.27
0.50
MS-Mapping [ 3 ]
0.26
0.45
0.29
0.40
Ours
0.17
0.38
0.24
0.47
TABLE II : Comparison of ATE RMSE (m) for multi-session approaches on Parkland (S1) and Short (S2) sequences.
Fig. 8 : Multi-session results in the Kemi underground mine. The boxed region shows a loop closure enforced between two areas previously mapped but disconnected. Furthermore, the zoomed-in images further illustrate the impact of global PGO.
Maintaining consistency over long spatial and temporal horizons remains a fundamental challenge in large-scale LiDAR SLAM, particularly when integrating maps collected across multiple sessions. We present Chain-SLAM, a LiDAR SLAM backend enabling online multi-session map alignment and reuse with global consistency at large scale. We implement a chained loop closure mechanism that efficiently propagates geometric constraints across inter-session keyframes through an adjacency graph, enabling robust long-horizon consistency triggered by reliable short-horizon loop closures. The system initializes inter-session alignment with GNSS-proximity place recognition, then performs on-the-fly loop closure detections and joint optimization of loaded maps and newly acquired trajectories within a unified factor graph, maintaining both inter- and intra-session geometric consistency without dynamic object removal, and cross-platform robustness with minimal hyperparameter tuning. Experimental results show improved trajectory accuracy and robust multi-session integration on large-scale datasets. We release our source code to support reproducible research in large-scale multi-session LiDAR SLAM. Project site: https://ai4ce.github.io/Chain-SLAM/
The simultaneous localization and mapping community has introduced a growing number of systems adapted for multi-session operations where the operational environment features low-dynamic changes that impact mapping, such as surface wear, weather phenomena, or seasonal change. These systems allow for lifelong operations by a robot within these environments. There is also growing interest in operations in environments where the unique ground texture is the only mapping feature available for use. These ground texture systems are not yet targeted for multi-session low-dynamic-change environments though. This work explores the impact of three different techniques on trajectory estimation accuracy in these multi-session low-dynamic ground texture environments. Of the three, the use of Kullback-Leibler Divergence, as a similarity score and a bias influencing loop closure confidence, is found to have the most success. We show an analysis of all three methods and a deeper exploration of the impact of Kullback-Leibler Divergence. We also introduce a dataset for use by the robotics community that contains multi-session images where the ground changes between sessions and also high-accuracy pose information for use in evaluation.
Kyle M. Hart, Brendan Englot
Naval Air Warfare Center, Aircraft Division, Lakehurst, NJ, 08733, USA
A conventional embodied agent in an indoor environment builds a map by accumulating noisy real-world measurements and results from heavy inference to guide navigation. While the detailed measurements may differ due to scene changes or sensor settings, the unique topological structure of navigable passages or the connectivity between rooms may provide training-free guidance that complements local observations. We propose a sparse topometric graph as an efficient structural representation for robust indoor navigation, alleviating the need for fine-grained matching, which is computationally expensive, memory-intensive, and prone to real-world variations. The map consists of nodes that encode local 3D geometry and distance-weighted edges that capture traversable connectivity. A learned graph-matching network finds consistent subgraph structures across sessions by considering both node-level geometric descriptors and neighborhood connectivity. The holistic contextual matching is robust to local ambiguities caused by weak distinctiveness or repetitive structure, and establishes stable correspondences despite local scene changes, different sensor modalities, or limited spatial overlap. We analyze the practicality of the graph through cross-session point-goal navigation, where a goal specified in a prior reference map must be reached from an unknown initial pose in a later session. Experiments show that our method achieves the highest success rates (77.1% on Gibson and 40.0% under 3D-FRONT scene changes), while reducing memory by more than two orders of magnitude compared to point-level representations. The proposed approach further supports cross-session alignment and goal-directed navigation with strong scalability and robust real-world transfer.