A possible thesis thread: robots that see the world through different sensors, on land, at sea and in the air, building one picture of it that a person can query and rely on.
A drone sees roofs and roads from above, a ground robot sees doors and kerbs, a boat sees a shoreline and a radar return. Each robot’s map lives in its own frame, uses its own labels and has its own errors, so merging them takes geometry, semantics and trust at once.
Scene graphs
Most groups now store maps as 3D scene graphs: objects, places and rooms as nodes, relations as edges, and language features on top (open-vocabulary mapping). Only Hydra-Multi and CURB-SG build one graph from several robots, and both assume a shared frame, honest teammates, good links and the same label set. The field’s own survey gives multi-robot work one sentence.
Sharing maps
What robots send each other keeps shrinking: whole graphs, then object maps, then labels and centroids, then fixed-size descriptors with images sent only on request. See collaborative SLAM.
Models and sensors
Foundation models give a cheap robot dense depth from one camera and segmentation by text prompt, along with domain shift and confident errors (foundation models for robotics). Outdoors there are no rooms to organise a map around, small drones carry 27 g of everything (aerial swarms), and under water even localisation is hard (underwater perception).
Language
Language works both ways: robots can describe what they see in it, and people can ask them things with it. See language-grounded planning and semantic coordination.
Open questions
- How do robots agree on a node when one says “chair”, another says “grey office chair” and a third has only a lidar cluster?
- Given several maps that disagree, which claims should the merged map keep? The scene graph explainer and the project on my research page start from this question.
- What is the smallest message that still lets a teammate query my map in language?
- Is agreement between a camera and a lidar worth more than agreement between two cameras running the same detector?
- How do you tie perception errors to task outcomes, when graph quality and task success are measured on different datasets?
Related published work: GoGet, which uses agreement between a segmenter and a vision-language model to decide when to trust a count.
Papers
| Paper | Year | Venue | Domain | Sensors |
|---|---|---|---|---|
| USS-Nav | 2026 | arXiv preprint | air | rgb-d, lidar, language |
| Wat3R | 2026 | ECCV 2026 | sea | camera |
| Fast-SAM3D | 2026 | ICML 2026 | general | camera |
| Pow3R-SLAM | 2026 | arXiv 2026 | land | rgb-d |
| MoGe-3 | 2026 | arXiv 2026 | general | camera |
| QuerySplat | 2026 | arXiv preprint | general | camera |
| Hickory | 2026 | arXiv preprint | land | rgb-d |
| Fast-FoundationStereo | 2026 | CVPR 2026 | general | stereo |
| InfiniDepth | 2026 | CVPR 2026 | general | camera, rgb-d |
| PSG-Nav | 2026 | ICML 2026 | land | rgb-d, language |
| LOST-3DSG | 2026 | arXiv preprint | land | rgb-d, language |
| DAAAM | 2026 | CVPR 2026 (per the code README; arXiv lists no venue) | land | rgb-d, language |
| TrackGraph | 2026 | arXiv 2026 | land, air | rgb-d, lidar, imu, language |
| Hydra++ | 2026 | IROS 2026 (accepted, per arXiv comments) | land | rgb-d, lidar, camera |
| SceneGraphVLM | 2026 | arXiv preprint | general | camera |
| BiMoSG | 2026 | arXiv preprint | land | rgb-d, language |
| 3DSG survey | 2026 | arXiv preprint | general | rgb-d, lidar, imu, language |
| vS-Graphs | 2026 | IEEE RA-L, vol. 11, no. 8, August 2026 | land | rgb-d |
| Picasso | 2026 | arXiv preprint | general | rgb-d |
| SuperMap | 2026 | RSS 2026 | land | lidar, camera, imu, language |
| CAPruner | 2026 | ACL 2026 | general | rgb-d, language |
| DRACo-SLAM2 | 2025 | arXiv preprint | sea | sonar, imu, dvl |
| Underwater enhancement and 3D review | 2025 | arXiv preprint | sea | camera, sonar |
| RUSplatting | 2025 | BMVC 2025 | sea | camera |
| Underwater multi-modal spatial perception | 2025 | ICRA 2025 Workshop on Field Robotics | sea | camera, stereo, imu, dvl |
| MERLION | 2025 | ICRA 2025 | sea | camera, language |
| RoboTwin 2.0 | 2025 | arXiv 2025 | land | camera, rgb-d, language |
| VSLAM-LAB | 2025 | IROS 2025 | general | camera, rgb-d, stereo, imu |
| MoGe | 2025 | CVPR 2025 | general | camera |
| SAM 3 | 2025 | arXiv 2025 | general | camera, language |
| UniDepthV2 | 2025 | arXiv 2025 | general | camera |
| MoGe-2 | 2025 | arXiv 2025 | general | camera |
| FunGraph | 2025 | IROS 2025 | land | rgb-d, language |
| Sea-thru distance accuracy | 2024 | ACRS 2024 | sea | camera |
| Opti-acoustic semantic SLAM | 2024 | arXiv preprint | sea | camera, sonar, imu, dvl, pressure |
| Gemma | 2024 | arXiv 2024 | general | language |
| CURB-SG | 2024 | ICRA 2024 | land | lidar |
| UniDepth | 2024 | CVPR 2024 | general | camera |
| SAM 2 | 2024 | arXiv 2024 | general | camera |
| OS-Atlas | 2024 | arXiv 2024 (cited as ICLR 2025) | general | camera, language |
| LSG | 2024 | arXiv preprint | land | rgb-d, lidar, imu |
| xFLIE | 2024 | arXiv preprint | land, air | rgb-d, lidar, imu |
| OpenGraph | 2024 | arXiv preprint | land | camera, lidar, language |
| Hydra (IJRR) | 2024 | The International Journal of Robotics Research (IJRR), 2024 | land | rgb-d, imu, lidar |
| Open3DSG | 2024 | CVPR 2024 | land | rgb-d, language |
| Clio | 2024 | IEEE RA-L 2024 | land | rgb-d, language |
| Khronos | 2024 | RSS 2024 | land | rgb-d, imu |
| HOV-SG | 2024 | RSS 2024 | land | rgb-d, lidar, language |
| DM_underwater | 2023 | arXiv 2023 | sea | camera |
| RT-2 | 2023 | CoRL 2023 (PMLR 229) | land | camera, language |
| Hydra-Multi | 2023 | IROS 2023 | land | rgb-d, imu, lidar |
| SAM | 2023 | ICCV 2023 | general | camera |
| Whisper | 2023 | ICML 2023 (arXiv 2022) | general | audio, language |
| LLM-guided scene graph planning | 2023 | arXiv preprint | land | language |
| PaLM-E | 2023 | ICML 2023 | land | camera, language |
| SayPlan | 2023 | CoRL 2023 | land | language, lidar, rgb-d |
| ConceptGraphs | 2023 | ICRA 2024 | land | rgb-d, language |
| 3D VSG | 2023 | ICRA 2023 | land | none |
| Common-sense spatial knowledge (workshop) | 2023 | RSS 2023 Workshop on Robot Representations for Scene Understanding, Reasoning and Planning | land | rgb-d |
| Language-enabled spatial ontologies | 2023 | IEEE RA-L 2024 | land | rgb-d, lidar, imu, language |
| Mono-Hydra | 2023 | ISPRS Annals (GSW 2023) | land, air | camera, imu |
| Embodied AI survey | 2022 | IEEE Transactions on Emerging Topics in Computational Intelligence 2022 (arXiv 2021) | land | camera, rgb-d, language |
| Taskography | 2022 | CoRL 2021 | land | none |
| S-Graphs | 2022 | IEEE RA-L 2022 (with IROS 2022) | land | lidar |
| Hydra | 2022 | RSS 2022 | land | rgb-d, stereo, imu |
| DSG navigation policies | 2021 | ICRA 2022 | land | stereo, imu |
| Kimera | 2021 | IJRR 2021 | land | stereo, rgb-d, imu |
| 3D Dynamic Scene Graphs | 2020 | RSS 2020 | land | stereo, imu |
| 3DSSG | 2020 | CVPR 2020 | land | rgb-d |
| 3D Scene Graph (Armeni) | 2019 | ICCV 2019 | land | camera, rgb-d |
| 3-D Scene Graph (Kim) | 2019 | IEEE Transactions on Cybernetics | land | rgb-d |
| Active perception | 2018 | Autonomous Robots 42(2), 2018 (arXiv 2016) | general | camera |
| Deep ensembles | 2017 | NeurIPS 2017 | general | camera |
| Transformer | 2017 | NeurIPS 2017 | general | language |
| Query by committee | 1992 | COLT 1992 (Proceedings of the Fifth Annual Workshop on Computational Learning Theory) | general | none |
| Paper | Year | Venue | Domain | Sensors |
|---|---|---|---|---|
| GS underwater study | 2026 | arXiv 2026 | sea | camera |
| AGT-CV | 2026 | ECCV 2026 Workshops (CDEL) | air, land | lidar, stereo, imu, gps, camera, thermal |
| Mapping Pamir | 2026 | ICRA 2026 | sea | camera, imu, pressure |
| Swimm3R | 2026 | arXiv 2026 | sea | camera |
| Benthic long-term localisation | 2026 | Frontiers in Robotics and AI 2026 | sea | stereo |
| Above and Below | 2026 | arXiv preprint | sea | sonar, lidar, imu, dvl |
| AquaFlow | 2026 | arXiv 2026 | sea | camera |
| MAG-VLAQ | 2026 | arXiv preprint | air, land | camera, lidar |
| AUSLUN | 2026 | arXiv preprint | sea, air | camera, imu, rangefinder |
| Sonar VFM adaptation | 2026 | arXiv preprint | sea | sonar |
| LeRobot | 2026 | arXiv 2026 | land | camera |
| MapAnything | 2026 | 3DV 2026 | general | camera, rgb-d |
| BatSLAM 2.0 | 2026 | arXiv preprint | land | sonar |
| LoGeR | 2026 | arXiv 2026 | general | camera |
| HEROIC | 2026 | arXiv preprint | air, land | camera, language |
| SGM-SLAM | 2026 | arXiv preprint | land | lidar, camera, imu |
| CERPE | 2026 | arXiv preprint | land | camera |
| Conformal map fusion | 2026 | arXiv preprint | general | none |
| PERSEPHONE perception | 2026 | ICARCV 2026 | land | lidar, camera, imu, language |
| CoSMO | 2026 | CoRL 2026 | land | camera, language |
| LTR² | 2026 | IEEE Transactions on Robotics (accepted) | land | lidar, radar |
| CILC | 2026 | arXiv preprint | general | camera, lidar |
| SABER | 2026 | arXiv preprint | land | lidar |
| Kairos | 2026 | arXiv preprint | land | camera |
| Outdoor open-set 3DSGs | 2026 | IEEE Transactions on Field Robotics (accepted) | land | camera, lidar, language |
| Mono-Hydra++ | 2026 | arXiv preprint | land | camera, imu |
| ChronoGraph | 2026 | arXiv preprint | land | camera, language |
| UQ-DAAAM | 2026 | arXiv preprint | land | camera, language |
| 3DSG semantic uncertainty | 2026 | arXiv preprint | land | camera, language |
| Underwater metric depth benchmark | 2025 | arXiv 2025 | sea | camera |
| WaterSplatting | 2025 | 3DV 2025 | sea | camera |
| Enhancement vs matching | 2025 | arXiv 2025 | sea | camera |
| UW-GS | 2025 | WACV 2025 | sea | camera |
| AQUA-SLAM | 2025 | IEEE T-RO 2025 | sea | stereo, imu, dvl |
| SeaSplat | 2025 | ICRA 2025 | sea | camera |
| SmolVLA | 2025 | arXiv 2025 | land | camera, language |
| SAM 3D | 2025 | arXiv 2025 | general | camera |
| MASt3R-SLAM | 2025 | CVPR 2025 | land | camera |
| CRISP | 2025 | CVPR 2025 | general | rgb-d |
| VGGT | 2025 | CVPR 2025 | general | camera |
| Depth Anything 3 | 2025 | arXiv 2025 | general | camera |
| ASHiTA | 2025 | CVPR 2025 | land | rgb-d, language |
| Multi-robot language-grounded planning | 2025 | arXiv preprint | land | rgb-d, lidar, language |
| Caustics removal via SLAM | 2024 | ECCV 2024 | sea | camera |
| Refractive COLMAP | 2024 | IROS 2024 | sea | camera |
| TURTLMap | 2024 | IROS 2024 | sea | stereo, imu, dvl, pressure |
| π0 | 2024 | RSS 2025 | land | camera, language |
| OpenVLA | 2024 | arXiv 2024 | land | camera, language |
| ROMAN | 2024 | RSS 2025 | land, air | rgb-d, imu |
| Eiffel Tower dataset | 2023 | IJRR 2023 | sea | camera |
| SeaThru-NeRF | 2023 | CVPR 2023 | sea | camera |
| C-3PO | 2023 | IEEE Transactions on Robotics 2023 (first arXiv version 2022, titled "Correct and Certify") | general | rgb-d |
| TEASER++ | 2021 | IEEE Transactions on Robotics 2021 | general | rgb-d, lidar |
| Neural Trees | 2021 | NeurIPS 2021 | general | none |
| GNC | 2020 | IEEE Robotics and Automation Letters 2020 (ICRA 2020 Best Paper Award in Robot Vision) | general | lidar, camera |
| Sea-thru | 2019 | CVPR 2019 | sea | rgb-d |
| PCM | 2018 | ICRA 2018 | land | lidar, imu |
air, land3
general36
land59
land, air4
sea28
Other threads: robust robotics, long-term autonomy. All three on one page: thesis map.