Papers for

drone navigation teams

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Hierarchical sparse graph enables accurate event based tracking across time

Hierarchical Sparse Graph for Multi-dimensional Assignment in Event-based Tracking

Abstract: Existing event-based tracking methods typically operate in an online manner, associating tracklets across adjacent batches. However, this limited temporal window often leads to failures in challenging scenarios, such as long-term occlusions or missed detections, where broader temporal context should be incorporated for high-precision tracking. To achieve robust and accurate tracking, this paper concentrates on the event-based multi-dimensional assignment that associates tracklets across multiple batches simultaneously. Unlike image-based detectors, event-based detectors can generate multiple correct yet spatially distinct tracklets for a single object per batch. This violates the one-to-one rule in classic data association, making event-based association an underdetermined many-to-many matching problem. To the best of our knowledge, this is the first work to tackle this NP-hard problem. To address its underdetermined nature, we formulate the assignment problem via fuzzy logic, where pairwise tracklet correlations are quantified through multiple long-range kinematic metrics. This fuzzy design eliminates the need for hard tracking thresholds and enables multiple heterogeneous metrics to be integrated into a unified correlation. By treating tracklets as nodes and their fuzzy correlations as edges, we construct a Hierarchical Sparse Graph (HSG). Building upon the HSG, we pose assignment as a graph partition task, and employ a heuristic solution that measures vertex affinity for partitioning adaptively. The entire pipeline is training-free. Comprehensive experiments on two challenging event datasets, eTram and Ev-UAV, show that our method outperforms existing event-based and most image-based association models. Particularly, it achieves over 2$\times$ the performance of OC-SORT across all metrics in the many-to-many association scenario.

Tue 27 OctComputer Vision and Pattern Recognition
The gist
Tracking objects that generate event data is hard, especially when objects disappear or get missed for a while. The authors handle this by linking object tracks over multiple time batches at once, instead of just nearby times. Their method treats the problem like connecting a fuzzy web of possibilities, allowing multiple matches without strict rules. This approach uses a special graph and smart grouping techniques to find the best matches without needing training. Tests show it works better than existing event and video tracking methods, especially in tough cases.
Open → 2610.37321v1

Embodied agents learn to handle task interruptions and resume navigation

Preemptive Instruction Following with Resumption

Abstract: Long-horizon embodied navigation may be interrupted by a higher-priority instruction before the current task sequence is complete. The agent must finish its current atomic task, execute the inserted task, and resume the remaining sequence from the updated environment state, where its viewpoint and available navigation candidates may have changed. To evaluate this capability, we introduce PIFR benchmark, a novel task for runtime task preemption and resumption in embodied navigation. PIFR includes single, repeated, and nested interruption settings,with stage-specific metrics for inserted-task completion, resumption of the original task, and completion of the full sequence. Evaluations show that recovery performance varies across interruption conditions, while aggregate success can obscure differences in interruption handling and resumption. We further propose a lightweight candidate-ranking baseline to improve target re-grounding after interruption. Together, PIFR and the baseline provide a systematic framework for studying interruption handling and task recovery in long-horizon embodied navigation.

Mon 26 OctRobotics
The gist
Sometimes robots or agents navigating a space get new orders before finishing their current task. This paper by the authors introduces a test called PIFR to see how well agents can pause their current task, do the new one, and then continue where they left off. They found agents work better in some interruption types than others, and simple methods can help them get back on track. This helps computers that move around follow instructions even when those instructions change suddenly.
Open → 2610.35576v1

Command aligned sampling improves drone motion planning efficiency

Extended LASMP: Language Aided Subset Sampling Based Motion Planner for Drone Navigation

Abstract: This paper proposes a command-aligned sampling distribution that incorporates navigation commands directly into the exploration process to improve the efficiency of sampling-based motion planning for quadcopters in three-dimensional environments. Inspired by the Language-Aided Subset Sampling Motion Planner, this paper extends command-aware subset sampling to aerial systems. Subset sampling is generalized to command-aligned regions, enabling structured 3D exploration while preserving probabilistic completeness. Natural language instructions are grounded into yaw-conditioned and altitude-aware geometric constraints using a prompt-based large language model for entity extraction and instruction validation. Extensive simulations in complex environments demonstrate that the proposed method generates collision-free, consistent trajectories while significantly reducing sampling effort compared to uniform RRT-based planners.

Sun 25 OctRoboticsArtificial Intelligence
The gist
Planning the movements of drones through complex 3D spaces is challenging and slow using usual methods. The authors propose a new approach that uses natural language commands, like directions, to guide how the path planning samples possible moves. This method improves efficiency by focusing on areas relevant to the commands while still ensuring safe, collision-free flight paths. They tested it extensively in simulations, showing it reduces the effort needed compared to traditional random methods.
Open → 2610.34054v1

Segment moving objects in video despite camera motion without training

Segmentation of Moving Objects During Camera Motion Using the OWL Function

Abstract: We present a principled, analytical framework for segmenting multiple independently moving rigid objects during single camera motion, built directly upon the OWL perception function. OWL encodes two fundamental visual motion cues, perceived looming $L$ and perceived rotation $ω$, into a single closed-form representation of relative 3D structure. A key consequence of this structure is that the unit direction of relative translation $\hat{t}$ is analytically recoverable per image point from $L$ and $ω$ alone. All points belonging to a non-rotating rigid body share the same $\hat{t}$, whereas points on different rigid bodies generally yield distinct values. This furnishes a natural segmentation criterion: clustering per-point $\hat{t}$ estimates partitions the visible scene into independently moving rigid bodies. Building on the 2D complex form of OWL, we derive the framework in 3D using quaternions, for a single moving camera. The approach is parallel, camera rotation-invariant, and requires no training data, no depth information, and no semantic priors.

Sat 24 OctComputer Vision and Pattern RecognitionRobotics
The gist
When a camera moves, it can be hard to tell which parts of the scene are objects moving on their own versus just shifting because of the camera. The authors use special visual clues about how objects look like they grow or rotate to figure out the direction each object is moving. By grouping parts of the image that move the same way, they can split the scene into separate moving objects. This method works without needing prior knowledge about the scene, depth measurements, or any learning from examples.
Open → 2610.33789v1

Compact transformer learns fast motion from event cameras and imu

Tight Learned Event-Inertial Odometry (TLEIO): a Transformer-Based Architecture for End-to-End Motion from Event Streams

Abstract: Event cameras provide microsecond temporal resolution, high dynamic range, and sparse asynchronous measurements, making them compelling for odometry under fast motion and challenging illumination. Yet, accurate event-based state estimation remains difficult: existing systems often use events as a complement to frames or rely on feature tracking, correspondence search, and optimization-heavy backends, limiting the efficiency promised by event sensing. We introduce Tight Learned Event-Inertial Odometry (TLEIO), a compact filter-based pipeline that learns short-window relative motion directly from voxelized event streams and tightly fuses the resulting constraints with IMU propagation. Its learned front-end, EventsFormer, predicts local displacements and per-axis uncertainty from event voxels in a single forward pass, while a stochastic-cloning EKF fuses these overlapping predictions with IMU propagation, producing metric trajectories without explicit feature tracks and pose optimization at inference time. The network is trained directly from pose displacements, without requiring event-flow labels. We evaluate TLEIO on the TartanAir/TartanEvent benchmark and on real DAVIS240c sequences. On TartanAir, TLEIO reduces average ATE from 2.26 m to 1.10 m over the IMU-only TLIO baseline, while achieving up to 48.9% lower runtime than optimization-based event odometry methods.

Fri 23 OctComputer Vision and Pattern Recognition
The gist
Measuring motion quickly and accurately is hard in fast or dark environments. Event cameras can help because they capture changes very fast and in high detail, but existing methods are complicated and slow. The authors created a system called TLEIO that uses a transformer network to understand short bursts of event camera data and combines that with motion sensor readings, making fast and accurate motion estimates without heavy calculations. It runs faster and improves accuracy compared to previous methods.
Open → 2610.33356v1

MatchCityLoc improves drone location using city shapes and angles

MatchCityLoc: Aerial Localization in LoD City Models via Surface Normal Matching

Abstract: Autonomous Unmanned Aerial Vehicle (UAV) navigation in urban environments requires precise pose estimation. Because Global Navigation Satellite System (GNSS) signals are often unreliable in cities, visual localization serves as a crucial complement. Within this domain, textureless Level-of-Detail (LoD) city models provide a lightweight and privacy-preserving substitute for heavy 3D textured meshes. However, localizing against abstract geometry is challenging. Existing pipelines typically score 2D silhouette overlaps, which can suffer from severe geometric ambiguities. We present MatchCityLoc, a framework that shifts this paradigm to direct feature matching from a noisy 6-DoF pose prior. By extracting surface normals from the query image using a zero-shot foundation model and matching them against rendered LoD normals, we establish explicit 3D-to-2D correspondences. This enables a Perspective-n-Point (PnP) pose solve, which we iteratively refine through a render-and-match loop. Experiments demonstrate strong pose recall and accurate translation and rotation among the evaluated off-the-shelf LoD-based methods with very few renderings. By bridging the modality gap between aerial imagery and texture-free geometry, MatchCityLoc enables accurate localization using lightweight, privacy-preserving city models.

Mon 19 OctComputer Vision and Pattern Recognition
The gist
Finding exact drone locations in cities can be tough because GPS signals often don't work well between tall buildings. The authors designed MatchCityLoc to help drones figure out where they are by comparing the shapes and angles of roofs and walls seen from above with simple city models that don't have detailed textures. Instead of just matching building outlines, their method looks at the directions surfaces face, making it easier to find the right spot. This approach works well even with lightweight city data that protects privacy.
Open → 2610.27110v1

Thermal UAVs find location using 2D satellite maps without GNSS

ScalLoc: Continuous Oblique Thermal UAV Geo-Localization with a 2D Satellite Map

Abstract: Cross-modal Thermal Geo-localization (TG) provides an illumination-robust positioning solution for Unmanned Aerial Vehicles (UAVs) in Global Navigation Satellite System (GNSS)-denied environments. However, existing thermal geo-localization methods mainly focus on near-nadir observations or depend on map-side elevation priors, limiting their applicability to continuous oblique thermal localization with only geo-referenced 2D satellite maps. To address this problem, we propose ScalLoc, a scale-aligned local-to-map localization framework for continuous oblique thermal UAV geo-localization. ScalLoc follows a reconstruction--refinement--anchoring--propagation design: it reconstructs local window geometry from thermal sequences, refines camera poses using track-based pose optimization, adaptively anchors each window to the satellite map, and propagates validated states to recover frame-level global trajectories. To support systematic evaluation, we construct Thermal-Oblique, a real thermal UAV sequence dataset containing 216 urban and rural day--night trajectories and 28,112 thermal frames with frame-level annotations. Extensive experiments demonstrate that ScalLoc reduces the mean localization error from 112.69 m to 26.97 m compared with the VGGT-only configuration, corresponding to a 76.07% reduction, and achieves the lowest frame-weighted mean position error among the compared methods. These results demonstrate the feasibility of continuous oblique thermal geo-localization using only lightweight 2D satellite maps, without requiring digital surface models or preconstructed 3D maps. Code and dataset are available at https://github.com/FloralHercules/ScalLoc.

Sat 10 OctComputer Vision and Pattern Recognition
The gist
UAVs often struggle to know their exact location when GPS signals are unavailable, especially at night or in tough environments. The authors developed ScalLoc, a way for drones with thermal cameras to determine their position continuously by matching their thermal images with standard 2D satellite maps. Unlike previous methods that needed special 3D maps or elevation information, ScalLoc works with just simple 2D maps. The authors also created a large real dataset of thermal images for testing, showing their method reduces location errors by over 75%.
Open → 2610.14866v1

Spatial context improves cross view geo localization accuracy

Learning Holistic Scene Layout against View Shifts in Cross-View Geo-Localization

Abstract: Cross-view geo-localization remains challenging due to drastic appearance variations and geometric distortions caused by viewpoint changes. Existing methods primarily rely on discriminative feature learning or explicit geometric alignment, yet often overlook an important observation: humans tend to match scenes by integrating the spatial context of multiple landmarks rather than relying solely on isolated local visual cues. Inspired by this, we propose a Spatial Context Modeling (SCM) module, which traverses feature maps along four orthogonal directions and models interactions among the resulting directional representations at each spatial position. By aggregating complementary directional observations, SCM constructs richer scene layout representations. Since reliable spatial context depends on robust visual features, we introduce a High-Frequency Information Guidance (HFIG) module that extracts stable structural cues from the frequency domain. By emphasizing edges and contours that are less sensitive to appearance variations, HFIG encourages the model to learn more robust representations. Extensive experiments demonstrate that our method achieves competitive performance on standard cross-view geo-localization benchmarks while exhibiting strong robustness under more challenging scenarios with unknown orientations and limited fields of view (FoVs).

Fri 9 OctComputer Vision and Pattern Recognition
The gist
Finding locations from images taken at very different angles is hard because pictures can look very different. The authors noticed that people recognize places by looking at the arrangement of multiple objects rather than just single details. They created a new method that looks at the layout of scenes by analyzing features along four directions to get a fuller picture. They also added a way to focus on stable edges and shapes to avoid confusion from changes in color or lighting. Their method works well in tests, even when the images are from unusual angles or limited views.
Open → 2610.13553v1

Structure aware matching improves drone thermal image localization

Structure-Aware Matching for Robust UAV Thermal-to-Satellite Geo-Localization

Abstract: UAV thermal-to-satellite geo-localization localizes UAVs by matching real-time thermal images against georeferenced satellite images, enabling autonomous navigation in GNSS-denied and low-visibility conditions. However, substantial cross-modal disparities in color, brightness and viewpoint hinder conventional appearance-based methods from establishing reliable correspondences. While existing approaches predominantly address global modality shifts, they overlook the invariant scene structures inherently shared between thermal and satellite views. To bridge this gap, we propose SAMF, a Structure-Aware Matching Framework that explicitly exploits stable cross-modal structural cues to improve UAV thermal geolocalization. Specifically, the Directional Structure Enhancement Module (DSEM) strengthens robust structural representations by capturing directional and low-frequency spatial patterns, and the Modality-Aware Affine Remapping Module (MARM) subsequently applies affine correction to these enhanced features, further promoting cross-modal alignment. Additionally, we introduce the Thermal-UAV-Paired benchmark derived from the Thermal-UAV dataset. Extensive experiments on both the Boson-nighttime and Thermal-UAV-Paired datasets show that SAMF achieves state-of-the-art performance. Ablation studies further validate the effectiveness and synergistic complementarity of DSEM and MARM. The code and dataset are released at: https://github.com/YuxinWang879/SAMF.git.

Fri 9 OctComputer Vision and Pattern Recognition
The gist
Finding a drone's location using heat-based images is hard because these images look very different from satellite photos. The authors created a new method called SAMF that focuses on matching stable shapes and lines in the images, rather than colors or brightness. This approach helps the drone figure out where it is even in tough, low-visibility situations. They also made a new dataset to test their method and showed it works better than previous ones.
Open → 2610.12835v1

Structure and semantics improve matching across drones and satellites

S$^3$Geo: Structure-Semantic Synergistic Learning for Cross-View Geo-Localization

Abstract: Cross-view geo-localization (CVGL) aims to estimate geographic locations by matching images captured from different viewpoints, such as drone and satellite views. Existing methods mainly rely on visual representations, but often fail to jointly model fine-grained structural correspondences and semantic priors, making them prone to confusion between visually similar but semantically different regions, and thus limiting robustness under large viewpoint variations. To address these challenges, we propose \textbf{S$^3$Geo}, a structure-semantic synergistic learning framework for cross-view matching. Specifically, we first introduce a Decoupled Query Pooling (DQP) module to extract a compact set of region-aware features from dense tokens, enabling explicit modeling of local structural patterns. We then design a query-level contrastive learning scheme with an optimal transport (OT)-based formulation to establish soft correspondences under cross-view spatial misalignment. Furthermore, we incorporate a Semantic Knowledge Distillation (SKD) strategy from a frozen CLIP teacher to transfer semantic priors and relational structures, thereby improving discrimination on hard negatives. By operating synergistically, the semantic priors provide robust contextual filtering, which guides the structural module to establish precise spatial alignments. Experiments on the University-1652 and SUES-200 datasets demonstrate that \textbf{S$^3$Geo} consistently outperforms state-of-the-art approaches without increasing inference complexity, validating the effectiveness of jointly modeling structural and semantic information for CVGL.

Thu 8 OctComputer Vision and Pattern Recognition
The gist
Sometimes it’s hard to know exactly where a photo was taken if the picture comes from a very different viewpoint, like a drone versus a satellite. The authors created a new way to help computers match these photos better by combining details about the shapes in the image with knowledge about what things are in the pictures. They use a special method to find matching parts even when the angles are different, and they also learn from a smart model that understands what objects usually look like. Their method works better than previous ones without making the computer slower.
Open → 2610.11608v1

Multimodal drone dataset enables better aerial perception across sensors

MultiFly: A Real-World Multimodal Aerial Dataset with Annotation-Efficient Label Transfer and Cross-Modal Semantic Consistency

Abstract: We introduce MultiFly, a real-world, low-altitude UAV dataset for semantic perception across RGB, thermal, LiDAR, and radar modalities. MultiFly provides 17,272 synchronized samples from four suburban scenes with frame-wise annotations for 15 semantic classes, together with calibration and GNSS-RTK/IMU measurements. To avoid costly and inconsistent modality-specific annotation, we propagate labels from only 115 manually annotated RGB images through shared geometric representations to all four modalities. This approach generates semantic labels for 17,157 additional RGB images, 17,272 thermal images, 840M LiDAR points, and 3.4M radar points. Transferred annotations achieve 89.93% average agreement with held-out manual annotations, and 90.94% average semantic consistency across all six modality pairs. We further establish semantic segmentation benchmarks for all four modalities, revealing distinct architectural behavior for dense LiDAR and sparse radar data. Taken together, MultiFly provides a scalable foundation for multimodal aerial perception and, to the best of our knowledge, the first public real-world low-altitude aerial benchmark that combines consistent frame-wise semantic annotations for RGB, thermal, LiDAR, and radar. Data at https://github.com/markus-42/multifly.

Wed 7 OctRoboticsComputer Vision and Pattern Recognition
The gist
Understanding scenes from drones is tricky because different cameras and sensors see the world differently. The authors created MultiFly, a large dataset with images and measurements from regular cameras, heat sensors, laser scanners, and radar on drones flying low over towns. They cleverly labeled a small number of images by hand and then used geometry to copy those labels to many other images and sensors, saving a lot of manual work. This dataset helps computers learn to recognize objects consistently in all these different types of data from drones.
Open → 2610.10359v1

Collision risk metrics compared and extended to handle uncertainty

Expressiveness, Equivalence, and Uncertainty in Velocity Obstacles and Closest Point of Approach Metrics

Abstract: Time to Closest Point of Approach (TCPA), Distance to Closest Point of Approach (DCPA), and Velocity Obstacles (VOs), are widely used to assess and mitigate collision risk in autonomous navigation, yet their relationship and behavior under uncertainty remain largely unexplored. Assuming perfect state information, we establish a relationship between these representations over finite and infinite prediction horizons and derive conditions under which they provide equivalent characterizations of collision risk. Under bounded uncertainty, we extend the Closest Point of Approach (CPA) metrics and VO to convex relative-state sets. We show that in this setting, independently computed TCPA and DCPA bounds lose the joint relationship required for VO membership, while uncertainty-aware VOs preserve this relationship through a set-valued representation of collision-inducing velocities.

Mon 5 OctRobotics
The gist
When autonomous robots or vehicles try to avoid crashes, they use measures like Time to Closest Point of Approach (TCPA), Distance to Closest Point of Approach (DCPA), and Velocity Obstacles (VOs). The authors studied how these measures relate to each other and when they give the same information if everything is perfectly known. They then looked at what happens when there is uncertainty about positions and speeds, showing that some traditional methods lose accuracy unless uncertainty is handled carefully. Their extended methods better represent collision risks when the exact state is uncertain.
Open → 2610.07371v1

Geometry semantic constraints improve ground camera localization with satellite images

How to Reduce Localization Ambiguity? Geometry-Semantic Constrained BEV Representation Learning for Satellite-Ground Localization

Abstract: Satellite-ground localization estimates the planar position and yaw orientation of a ground camera within a geo-referenced satellite image. Most recent methods map ground and satellite features into a shared bird's-eye-view (BEV) space and establish spatial correspondences. However, insufficient depth constraints can assign one ground feature to different distances along a viewing direction, creating geometric ambiguity in BEV feature placement. Similar appearances at different locations can also create descriptor matching ambiguity, while existing descriptor learning lacks explicit semantic supervision to distinguish them. We propose GeoSem-BEV, a geometry-semantic constrained BEV representation learning method. Radial depth supervision constrains distance assignment, and vertical height supervision constrains height aggregation. Shared explicit semantic supervision promotes consistent semantic predictions across views and helps distinguish locations with similar semantics. These constraints improve feature placement and descriptor discriminability, enhancing state-of-the-art BEV localization models. On VIGOR with unknown orientation, GeoSem-BEV reduces mean orientation error by 37.2% and 38.1% in the cross-area and same-area settings, respectively. The corresponding errors are reduced by 10.8% and 15.6% on DReSS-D. On KITTI-CVL, it reduces same-area mean orientation error by 26.8% under 10 degree orientation noise.

Wed 30 SeptComputer Vision and Pattern Recognition
The gist
Satellite-ground localization finds where a ground camera is positioned within a satellite image, including which way it is facing. The problem is that features can appear similar or be placed at uncertain depths, causing confusion about exactly where the camera is. The authors propose a method called GeoSem-BEV that uses geometric rules about depth and height, plus semantic understanding of the scene, to better place features and tell them apart. This approach helps reduce errors in determining the camera’s location and orientation when tested on multiple datasets.
Open → 2609.39127v1

Coupled 3D mapping method improves navigation planning and obstacle detection

Planning Oriented 3D Scene Completion via Coupled TUDF Occupancy Representation Learning from Partial Observations

Abstract: Partial observability remains a fundamental challenge in robotic navigation, where limited sensor coverage and occlusions leave large portions of the environment unobserved. Existing scene completion methods primarily focus on improving incomplete mapping or reconstructing partially observed 3D structures, but rarely investigate how scene completion can be designed to benefit downstream tasks such as path planning. In this work, we propose a path-planning-oriented 3D scene completion framework that moves beyond pure occupancy modeling toward a coupled geometric formulation. Specifically, given partial LiDAR observations as input, the proposed framework jointly predicts completed Truncated Unsigned Distance Field (TUDF)-based continuous geometric representations and voxel-wise occupancy maps. This coupled representation allows the network to better reason about obstacle boundaries and free-space geometry. To fully exploit the synergy between the two representations, we introduce a bidirectionally coupled learning scheme, where TUDF features provide dense geometric guidance to improve occupancy reconstruction, while occupancy features in turn offer complementary structural constraints that refine distance-field estimation. Consequently, the proposed network directly predicts complete occupancy and TUDF representations, allowing seamless integration of TUDF into trajectory planning without post-processing. Extensive experiments on unseen environments demonstrate that the proposed method consistently improves both geometric reconstruction quality and downstream planning performance.

Tue 29 SeptRobotics
The gist
Robots often have trouble navigating because their sensors can't see the entire environment due to obstacles or limited range. The authors developed a method that helps robots fill in the missing parts of a 3D map by predicting both where obstacles are and the detailed shapes of free spaces. Their approach combines two ways of representing the environment to make the predictions better and directly useful for planning safe paths. Tests showed this method improves both the accuracy of mapping unknown areas and helps robots plan routes more effectively.
Open → 2609.36543v1

Visual place recognition improves with reliability guided aggregation

Calibrating Retrieval Geometry: Reliability-Guided Training-Free Aggregation for Visual Place Recognition

Abstract: Frozen visual foundation models provide transferable features for visual place recognition, but fixed aggregation can suppress useful distinctions in new environments. We introduce TFA, a reliability-guided, training-free aggregation method requiring neither place labels nor task-specific weight updates. Our key observation is that reproducible retrieval need not be discriminative: independent codebooks can consistently retrieve a few database hubs. TFA combines cross-codebook agreement, retrieval coverage, and spectral statistics to control residual assignment, spectral shaping, and global-feature fusion. Its spectral kernel exactly recovers original descriptor similarity at zero intervention. Database-only TFA fixes its rules before accessing queries; TFA-C64 uses 64 disjoint unlabeled target images to calibrate retrieval for subsequent queries. Across 20 ground protocols with a fixed DINOv2-B backbone and matched resolution, database-only TFA improves Recall@1 over AnyLoc by 17.39 percentage points on MSLS-val and 9.55 on SPED. C64 mitigates failures of database-only calibration in driving environments. Across eight aerial/cross-view protocols, TFA achieves the highest Recall@1 among compared training-free heads in 14 of 16 DINOv2/DINOv3 backbone-protocol combinations. In a separate native-system comparison, DINOv2-G-based TFA-C64 reaches 91.46% Recall@1 on Pitts30k and 76.29% on VPAIR, outperforming the displayed training-free comparators on all five benchmarks. These results show that reliability-guided aggregation can recover additional retrieval capability from frozen representations, providing a practical baseline for new environments with scarce place supervision.

Tue 22 SeptComputer Vision and Pattern Recognition
The gist
Visual place recognition helps computers identify where a photo was taken by comparing image features. The authors found that fixed methods for combining these features can miss important details in new environments. They created a method called TFA that groups features reliably without needing extra training or labels. This method adjusts based on how consistent different feature sets are and can improve recognition accuracy in various challenging settings.
Open → 2609.25937v1

UAVs use predicted views to explore hidden spaces faster and better

WOLF: World Model Guided LiDAR Exploration with Predictive Frontiers

Abstract: LiDAR-based unmanned aerial vehicle (UAV) exploration builds maps by continually selecting where to observe next. However, decisions based on the measured map provide limited foresight into spatial continuations behind occlusions, leaving potentially informative directions unrecognized. We present WOLF, a world-model-guided framework that predicts future observations to enhance autonomous exploration. In the training stage, a recurrent world model learns observation dynamics from exploration trajectories, with recurrent memory retaining the spatial context needed to interpret partial observations across successive views. Building on this context, the model combines observation history with candidate motions during exploration to predict local occupancy and visibility. To guide further sensing, a predictive frontier generation mechanism then aligns and fuses these predictions using confidence, branch agreement, and observation quality to identify promising regions. The resulting predictive frontiers join measured ones to guide geometric viewpoint selection and trajectory generation, while new scans update subsequent predictions. In simulations, our method reduces mean terminal time by 10.9% relative to EPIC in Garage at comparable coverage and increases mean coverage from 42.12% to 98.35% in Tunnel. Real-world experiments further demonstrate onboard deployment of the learned model for online inference during physical flight.

Sun 20 SeptRobotics
The gist
Mapping unknown places with drones using laser sensors can miss important spots hidden behind obstacles. The authors present WOLF, a method where a smart model predicts what is likely to be behind these obstacles based on past observations, helping the drone decide where to look next. This prediction guides the drone more effectively, speeding up exploration and covering more area. They tested WOLF in simulations and real flights, showing improved speed and completeness compared to a prior method.
Open → 2609.23656v1

PerSeM improves long-term semantic mapping for UAVs with persistent memory

PerSeM: Persistent Semantic Memory for Long-Horizon Open-Vocabulary UAV Mapping

Abstract: Open-vocabulary segmentation enables rich semantic perception for UAVs, but frame-wise predictions can remain temporally inconsistent across repeated observations and changing viewpoints. We present PerSeM, a training-free persistent semantic memory framework for long-horizon open-vocabulary UAV mapping. PerSeM associates frame-wise semantic observations with persistent world-space voxels and constructs a majority-based semantic memory, which is conservatively refined through history-preserving spatial refinement, trust-aware replay, and context-guided verification. Experiments on the Forest and UAVScenes benchmarks show that persistent 3D memory provides substantial gains in semantic correctness and temporal stability over frame-wise predictions. Beyond this strong persistent-memory baseline, PerSeM provides consistent additional improvements, improving both semantic accuracy and temporal stability across all five evaluated UAVScenes sequences. Analysis using regions identified independently of the final PerSeM predictions further shows that these gains are concentrated in semantically difficult and temporally unstable regions, where majority-based memory is most likely to remain uncertain. These results demonstrate that persistent 3D aggregation provides a strong foundation for long-horizon semantic mapping, while conservative refinement of uncertain memory states can provide additional improvements without retraining or additional neural-network inference.

Thu 17 SeptComputer Vision and Pattern RecognitionRobotics
The gist
UAVs can see and label things around them but often get inconsistent results when looking repeatedly from different angles. The authors developed PerSeM, which keeps a memory of what the UAV has seen over time, combining many observations to get more accurate and stable labels. This method doesn’t need to be trained again and improves the UAV’s understanding especially in tricky places. Tests showed PerSeM makes the UAV’s semantic maps both more accurate and reliable over long periods.
Open → 2609.19542v1

Uncertainty-aware AI improves off-road robot navigation routes

UDAV: Uncertainty-Driven Adaptive VLM Waypoint Planner

Abstract: Vision-language models (VLMs) can generate routes directly from aerial imagery for off-road navigation, but their predictions provide no indication of reliability. We present UDAV, an Uncertainty-Driven Adaptive VLM Waypoint Planner for UAV-guided UGV navigation. UDAV draws multiple stochastic trajectory predictions, selects their medoid as a self-consistent nominal route, and estimates predictive uncertainty from their spatial dispersion. When the maximum uncertainty across interior waypoints exceeds a threshold, UDAV invokes a reconsideration stage; otherwise, it returns the medoid directly. We evaluate UDAV on 400 held-out trajectory queries from two UAV flights. Stochastic medoid selection reduces the mean average displacement error (ADE) from 147.4 pixels for a deterministic prediction to 115.9 pixels. The complete planner achieves a mean ADE of 110.4 pixels, a 25.1% reduction relative to deterministic planning, while producing valid trajectories for all queries. UDAV also yields the lowest 90th- and 95th-percentile errors among all evaluated configurations, including a higher-budget K=10 consensus baseline. Relative to the K=5 medoid, UDAV reduces these errors from 225.3 and 326.0 pixels to 199.0 and 290.8 pixels, respectively. These results demonstrate that stochastic VLM predictions provide both a stronger nominal route and an actionable uncertainty signal for selectively mitigating large planning errors.

Mon 14 SeptRoboticsArtificial Intelligence
The gist
Navigation systems for robots using aerial images can get confused and make mistakes without knowing how confident they are. The authors created a method called UDAV that makes many route guesses, finds the most typical one, and measures how reliable it is. If the route looks uncertain, UDAV rethinks the plan to avoid big errors. Tests show this idea makes navigation more accurate and safer for off-road vehicles guided by drones.
Open → 2609.16368v1