cs.ROOct 24, 2026

Memory-Aware Keyframe Management for Dense Maps on Memory-Constrained Robots

Authors: Huichang Yun, Jaeyeul Kim, Seungho Yoo

Organizations: Dept. of Computer Engineering, Pukyong National University, Busan, Korea

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Running a language model and a dense 3D map on one device means the two share one memory, and exceeding its limit is an out-of-memory failure. Map systems that keep a bounded working set bound it by a count of keyframes, but the memory one keyframe occupies varies widely with what it carries, so a count does not control what the working set occupies. The unit of admission is as poorly matched, since a radius around the pose, a fixed cell, or the graph neighbors of the last matched keyframe all move parts of the map that are not needed. We therefore propose two decisions. First, the memory the resident map uses is bounded directly, and the cost of every admission is known before any data is read, from per-keyframe resident sizes computed offline and carried with the map. Second, keyframes are admitted and released at a semantic spatial unit, the rooms and corridors of the building, so that the robot holds the zone it occupies together with the zone it is about to enter while the zones it has passed are released. Every admission measures the resident total, makes room, and only then reads, so the budget is never overshot. On two NVIDIA Jetson boards the budget holds in every run and the model gives up at most 0.8 percent of its throughput, whereas without it the operating system kills the model or the map. Under that budget, and over the environments evaluated, admission by semantic zone localizes 6-9 points more of the query frames than a geometric partition of matched zone count and about twice as much as the retrieval of RTAB-Map, at a lower ATE. Bounding memory lets the map and the model coexist, and moving the map in semantic zones makes the bounded map usable.

Figures & tables

Fig. 1 : Zone-granular residency under a memory budget, the concept of ZEBRA
Fig. 2 : Overview of ZEBRA
Fig. 3 : Mission routes R1–R3 on the two maps.
Fig. 4 : The three partitions, hospital left and office right.
DeviceModelModel sizeMapKeyframesPayloadResidual
  (GB) (#)(MB)(MB)
AGX Orin, 32 GBQwen3.5 27B25.0Hospital162083991750
AGX Orin, 32 GBQwen3.5 27B25.0Office104152632100
Orin Nano, 8 GBGemma3 1B ∗2.6Hospital162083992400
Orin Nano, 8 GBGemma4 E2B4.2Office10415263500
TABLE I : Devices, models, maps, and the residual budget given to the map, the memory the model leaves free. Model size is the resident size of the model. ∗ Text-only.
 AGX Orin, hospitalAGX Orin, officeOrin Nano, hospitalOrin Nano, office 
Map budget (MB)170016501600205020001950235023002250450400350Overall ↑\uparrow
Org: RTAB-Map45.131.134.35.97.37.739.130.515.06.88.513.920.4
Org: RTAB-Map, route34.742.331.05.816.17.328.724.330.38.07.96.220.2
ZEBRA: reference44.843.244.248.143.142.147.748.046.732.219.816.539.7
ZEBRA: VLM partition50.649.447.541.943.437.656.454.855.023.722.620.542.0
ZEBRA: cubic chunk26.227.324.844.443.641.636.233.235.927.533.526.633.4
TABLE II : Localization rate with respect to the unit of retrieval, mean over the three routes, percent of query frames with a pose.
Table 7
Table 8
More in this topic

3D Mapping

See all 36 papers

Explore similar work

CardsList
Mar 31, 2026

Semantic Zone-Based Map Management for Stable AI-Integrated Mobile Robots

Recent advances in large AI models (VLMs and LLMs) and joint use of the 3D dense maps, enable mobile robots to provide more powerful and interactive services grounded in rich spatial context. However, deploying both heavy AI models and dense maps on edge robots is challenging under strict memory budgets. When the memory budget is exceeded, required keyframes may not be loaded in time, which can degrade the stability of position estimation and interfering model performance. We proposes a semantic zone-based map management approach to stabilize dense-map utilization under memory constraints. We associate keyframes with semantic indoor regions (e.g., rooms and corridors) and keyframe management at the semantic zone level prioritizes spatially relevant map content while respecting memory constraints. This reduces keyframe loading and unloading frequency and memory usage. We evaluate the proposed approach in large-scale simulated indoor environments and on an NVIDIA Jetson Orin Nano under concurrent SLAM-VLM execution. With Qwen3.5:0.8b, the proposed method improves throughput by 3.3 tokens/s and reduces latency by 21.7% relative to a geometric map-management strategy. Furthermore, while the geometric strategy suffers from out-of-memory failures and stalled execution under memory pressure, the proposed method eliminates both issues, preserving localization stability and enabling robust VLM operation. These results demonstrate that the proposed approach enables efficient dense map utilization for memory constrained, AI-integrated mobile robots. Code is available at: https://github.com/huichangs/rtabmap/tree/segment

Huichang Yun, Seungho Yoo

Dept. of computer Engineering, Pukyong National University, Busan, Korea

Mobile RoboticsSimultaneous Localization and MappingSemantic MappingRobot Localization
AskPDF
Apr 13, 2026

A Survey of Spatial Memory Representations for Efficient Robot Navigation

As vision-based robots navigate larger environments, their spatial memory grows without bound, eventually exhausting computational resources, particularly on embedded platforms (8-16GB shared memory, <<30W) where adding hardware is not an option. This survey examines the spatial memory efficiency problem across 88 references spanning 52 systems (1989-2025), from occupancy grids to neural implicit representations. We introduce the α=Mpeak/Mmapα= M_{\text{peak}} / M_{\text{map}}, the ratio of peak runtime memory (the total RAM or GPU memory consumed during operation) to saved map size (the persistent checkpoint written to disk), exposing the gap between published map sizes and actual deployment cost. Independent profiling on an NVIDIA A100 GPU reveals that αα spans two orders of magnitude within neural methods alone, ranging from 2.3 (Point-SLAM) to 215 (NICE-SLAM, whose 47,MB map requires 10GB at runtime), showing that memory architecture, not paradigm label, determines deployment feasibility. We propose a standardized evaluation protocol comprising memory growth rate, query latency, memory-completeness curves, and throughput degradation, none of which current benchmarks capture. Through a Pareto frontier analysis with explicit benchmark separation, we show that no single paradigm dominates within its evaluation regime: 3DGS methods achieve the best absolute accuracy at 90-254,MB map size on Replica, while scene graphs provide semantic abstraction at predictable cost. We provide the first independently measured αα reference values and an αα-aware budgeting algorithm enabling practitioners to assess deployment feasibility on target hardware prior to implementation.

Ma. Madecheen S. Pangaliman, Steven S. Sison, Erwin P. Quilloy +1

University of the Philippines Diliman · University of Santo Tomas

Robotic Spatial MemoryRobot NavigationSimultaneous Localization and Mapping
AskPDF
Aug 7, 2026

M2-SMap: Memory-Efficient Semantic Mapping with Hierarchical Multi-Model Representation

Dense point cloud maps, as a typically used mapping representation, are difficult to deploy on resource-constrained robots because their memory consumption grows rapidly with scene scale. Although compact single-model representations reduce memory cost, their fixed geometric expressiveness is insufficient for structurally diverse environments. Existing multi-model methods improve representational flexibility, yet their feature extraction and model selection are often dominated by local geometry, which can cause overfitting and adhesion between objects. To address these issues, this paper presents M2-SMap, a memory-efficient semantic mapping framework based on hierarchical multi-model representation. First, a hierarchical geometric decomposition partitions RGB-D point clouds into compact Gaussian components. Then, a projection-guided semantic annotation mechanism assigns instance identities to each component. Subsequently, these annotations are incorporated into an object-aware Gaussian fusion strategy. Furthermore, a multi-scale feature extraction strategy separates large planar regions, semantic objects, and complex residual structures, which are respectively represented by bounded planes, object-level superquadrics, and GMM primitives. Experiments on three RGB-D sequences show that M2-SMap runs in real time at no less than 29.37 Hz while achieving the lowest primitive count, with an average reduction of 18.7% over the best baseline. It also reduces the mean per-frame number of measured inter-object adhesion cases from 2.808 to 0, demonstrating efficient and semantically consistent scene representation.

QiYing Deng, ZhongLai Wang, Yuan Gao +1

University of Electronic Science and Technology of China · State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, China

Superquadric Representation3D Scene RepresentationSemantic Mapping3D Semantic Mapping
AskPDF