MMUSV-Sim: A Maritime Simulation Platform for Multi-USV Cooperative Perception
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: We're looking at the paper: "MMUSV-Sim: A Maritime Simulation Platform for Multi-USV Cooperative Perception".
Dev: Here's the gist of the paper. Cooperative perception among multiple unmanned surface vehicles (USVs) combines complementary observations to extend maritime target sensing beyond the view range and field of a single platform.
Rosa: MMUSV-Sim is a perception-oriented maritime simulation and data-generation platform built on Unreal Engine 5 and Project AirSim that provides island, open-sea, and port environments with configurable weather, time of day, wave conditions, a diverse vessel asset library, and spline-based multi-vessel motion.
Dev: Experiments verify that the configured wave settings produce the intended changes in vessel heave, roll, and pitch.
Rosa: Platform Architecture MMUSV-Sim integrates configurable maritime scenario generation with state-consistent multimodal acquisition and shared world-state annotation generation.
Paper summary: Rosa: So, wrapping up this discussion on MMUSV-Sim and its platform for multi-USV cooperative perception, we’re looking at the core claims made by Ziao Li and her team <ref:2608.14207#pg1>. They developed a simulation platform that integrates configurable maritime environments, perception-oriented vessel motion, state-consistent multimodal acquisition, and per-agent annotation generation <ref:2608.14207#pg2>.
Dev: And the main results they show are pretty concrete. They confirm that configured wave levels produce monotonic increases in heave, roll, and pitch RMS across the tested settings <ref:two thousand six hundred eight point one four two zero
Conclusion: Rosa: So we’re wrapping up our look at MMUSV-Sim, which is this platform for multi-USV cooperative perception, and we need to talk about what that title actually means for us out here in the field.
Dev: It’s a simulation tool <ref:2608.14207#pg1> built on Unreal Engine five and Project AirSim, designed to let you test how multiple unmanned surface vehicles work together to sense things beyond what any single boat can see.
Taro: So it’s not just about one boat seeing a target better, it’s about combining all the observations from several different USVs to get a much bigger picture of the maritime environment.
Rosa: Exactly, and what this paper shows is that they’ve put together something that lets you configure these complex maritime scenarios—think island environments, open sea, or busy ports—and then generate data from it in a consistent way.
Dev: The authors describe a whole pipeline: they build the simulation engine to handle rendering and motion, then they have this module to set up the scenario and orchestrate everything that happens when you run an experiment.
Taro: And for the real world, what matters is how accurate that data generation is. They show that if you configure wave conditions correctly, you can actually see measurable changes in how much heave, roll, and pitch those vessels experience.
Rosa: That’s a big deal because it means we aren't just guessing about the physics; we can simulate the effect of different sea states on the vehicles themselves.
Dev: And they show that when you run these experiments, you get really clean results for things like detecting objects using LiDAR from those cooperative views.
Taro: The numbers they give on detection accuracy improvements under early fusion conditions are pretty strong, showing how much better it gets compared to just having one vehicle look around.
Rosa: It suggests that if we can build these kinds of platforms, the next step is taking this off the screen and seeing what happens when real USVs actually try to use those cooperative views in a messy environment.
Ziao Li, Jianxiong Ye, Biao Tang, Leping Zhang, Kun Zuo, Siyu Huang, Ánqiang Gao
School of Intelligent Systems Engineering, Sun Yat-sen University · Á School of Intelligent Imagery Engineering, Beijing Film Academy
cs.RO, cs.CV
Submitted: 2026-08-14
Updated: 2026-11-04
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 79/100
The gist: Cooperative perception among multiple unmanned surface vehicles (USVs) combines complementary observations to extend maritime target sensing beyond the view range and field of a single platform.
Key concepts
- Cooperative Perception
- This involves multiple USVs sharing their sensor data to create a more complete picture of the environment than any single vehicle could achieve alone. By combining viewpoints, the system can sense targets outside the range of any one platform.
- MMUSV-Sim
- This is a simulation platform using Unreal Engine 5 that generates realistic maritime scenarios. It allows users to configure detailed environments like islands or ports, control weather and waves, and model the motion of multiple USVs in synchronized ways for testing perception systems.
- Perception-Oriented Motion Model
- This model separates the planned route of a vessel from small, controllable movements caused by external factors like waves. It calculates where a vessel is based on its intended path and then adds synthetic up/down (heave) and side-to-side (roll) movements caused by wave conditions.
- Early Fusion
- This technique is used to combine sensor data from different USVs more effectively for detection. In experiments, using Early Fusion with LiDAR data improved the performance metric (AP@0.5) significantly compared to using only one vehicle's sensors.
Terminology
Summary
Cooperative perception among multiple unmanned surface vehicles (USVs) combines complementary observations to extend maritime target sensing beyond the view range and field of a single platform. p. 0 MMUSV-Sim is a perception-oriented maritime simulation and data-generation platform built on Unreal Engine 5 and Project AirSim that provides island, open-sea, and port environments with configurable weather, time of day, wave conditions, a diverse vessel asset library, and spline-based multi-vessel motion. p. 0 Experiments verify that the configured wave settings produce the intended changes in vessel heave, roll, and pitch. p. 0
Platform Architecture
MMUSV-Sim integrates configurable maritime scenario generation with state-consistent multimodal acquisition and shared world-state annotation generation. It consists of three core components: a UE5-based maritime simulation engine, a scenario configuration and orchestration module, and a control and data acquisition interface. p. 1 The UE5-Based Maritime Simulation Engine adopts UE5 as the simulation backbone for maritime scene rendering, multi-vessel instantiation, sensor mounting, and perception-oriented data generation. p. 1 This engine provides parameterized control over time of day, weather, illumination, and wave settings within a common simulation framework. p. 1
Scenario Configuration and Motion Modeling
MMUSV-Sim supports two complementary modes for constructing executable maritime scenarios: manual configuration for fine-grained scenario design and batch generation for automatically producing multiple executable configurations. In manual mode, users place USVs and target vessels and define their spline trajectories through a bird’s-eye-view (BEV) configuration interface. p. 2 The perception-centric vessel motion model separates configuration-driven route motion from controllable local pose perturbations. p. 2 This model uses a formulation to compute the horizontal position and yaw of each vessel based on its spline trajectory and nominal speed. p. 2 Synthetic heave, roll, and pitch perturbations are superimposed on the route motion using configurable wave settings. p. 3
Data Acquisition and Annotation Pipeline
The platform implements an acquisition and annotation pipeline that records sensor observations while scene evolution is paused. At each acquisition index, the interface pauses scene evolution while keeping the sensor streams active. p. 3 After a configured buffer-refresh interval, it records the refreshed multi-view camera, LiDAR, and radar outputs and queries the current USV poses and target-vessel states. p. 3 Offline post-processing generates per-agent annotations from shared world frames. p. 1
Evaluation of Cooperative Perception
The experiments evaluate three capabilities: controllable vessel-pose perturbations under configurable wave conditions, agreement between projected annotations and semantic renderings, and downstream utility for LiDAR-based cooperative BEV detection. The pose perturbation experiment shows that heave RMS increases monotonically from W0 to W3 across four wave settings. p. 6 Regarding annotation rendering agreement, the median center error between the projected box center and the centroid of the overlapping region is 2. p. 508 pixels across all scenes. p. 0 For LiDAR-based cooperative BEV detection, Early Fusion improves AP@0. p. 6 p. 05 from 45.54 to 72.74 relative to the ego-only baseline under clean conditions. p. 6 Fixed-delay and pose-noise experiments further characterize degradation under stale and misaligned cooperative observations. p. 6 p. 6
Conclusion
MMUSV-Sim integrates configurable maritime environments, perception-oriented vessel motion, state-consistent multimodal acquisition, shared world-state capture, and per-agent annotation generation. Configured wave levels produce monotonic increases in heave, roll, and pitch RMS across the tested settings. p. 6 Early Fusion improves AP@0.5 from 45.54 to 72.74 relative to the ego-only baseline. p. 6 Fixed-delay and pose-noise experiments further characterize degradation under stale and misaligned cooperative observations.
REFERENCES
[1] J. Kim, C. Lee, D. Chung, and J. p. 6 Kim, “Navigable area detection and perception-guided model predictive control for autonomous navigation in narrow waterways,” IEEE Robotics and Automation Letters, vol. p. 6 8, no. 9, pp. 5456–5463, 2023 <ref:2608. p. 614207#pg2>
[2] T. Clunie, M. DeFilippo, M. p. 6 Sacarny, and P. p. 6 Robinette, “Development of a perception system for an autonomous surface vehicle using monocular camera, lidar, and marine radar,” in 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. p. 6 14 112–14 119
[3] B. Bovcon, J. Muhovic, J. p. 6 Per s, and M. p. 0 Kristan, “The mastr t r 1325 dataset for training deep usv obstacle detection models,” in 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). p. 6 IEEE, 2019, pp. p. 6 3431–3438
[4] A. Nanda, S. W. Cho, H. Lee, and J. p. 6 H. Park, “Kolomverse: Korea open large-scale image dataset for object detection in the maritime universe,” IEEE Transactions on Intelligent Transportation Systems, vol. p. 6 25, no. 12, pp. 20 832–20 840, 2024
[5] M. Jeong, A. Chadda, Z. Ren, L. Zhao, H. Liu, A. Zhang, Y. Jiang, S. Achong, S. Lensgraf, M. p. 6 Roznere, and A Q Li “Seepersea: Multimodal perception dataset of in-water objects for autonomous surface vehicles,” IEEE Transactions on Field Robotics, vol 2 pp 737–752
[6] S. Yao, R. Guan, Y. Ni, S. Xu, Y. Yue, X. Zhu, and R W Liu “Usvtrack: Usv-based 4d radar-camera tracking dataset for autonomous driving in inland waterways,” in 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025, pp. p. 6 19 016–19 023
[7] A. Dosovitskiy, G. p. 6 Ros, F. Codevilla, A Lopez, and V Koltun “CARLA: An open urban driving simulator,” in Proceedings of the 1st Annual Conference on Robot Learning, 2017, pp 1–16
[8] R. Xu, Y. Guo, X. Han, X Xia, H Xiang, and J Ma “Opencda: An open cooperative driving automation framework integrated with co-simulation,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC), 2021, pp 1155–1162
[9] R. Xu, H Xiang, X Xia, X Han, J Li, and J Ma “Opv2v: An open benchmark dataset and fusion pipeline for perception with vehicle-to-vehicle communication,” in 2022 International Conference on Robotics and Automation (ICRA). p. 6 IEEE, 2022, pp. p. 6 2583–2589
[10] Y. Li, D Ma, Z An, Z Wang, Y Zhong, S Chen, and C Feng “V2xsim: Multi-agent collaborative perception dataset and benchmark for autonomous driving,” IEEE Robotics and Automation Letters, vol 7 no 4 pp 10 914–10 921
[11] F Han, L Zhang, X Wang, K-A Zhao, Y Zhong, Z Su, T Feng, and W Zhu “U2usim-a uav telepresence simulation platform with multi-agent sensing and dynamic environment,” in Proceedings of the 32nd ACM International Conference on Multimedia, 2024, pp 11 258–11 260
[13] N Koenig and A Howard “Design and use paradigms for gazebo, an open-source multi-robot simulator,” in 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), vol. 3, 2004, pp. p. 6 2149–2154
[14] M M M Manhaes, S A Scherer, M Voss, L R Douat, and T Rauschenbach “Uuv simulator: A gazebo-based package for underwater intervention and multi-robot simulation,” in OCEANS 2016 MTS/IEEE Monterey, 2016, pp. 1–8
[15] M Grimaldi, P Cieslak, E Ochoa, V Bharti, H Rajani, I Carlucho, M Koskinopoulou Y R Petillot and N Gracias “Stonefish: Supporting machine learning research in marine robotics,” in 2025 IEEE International Conference on Robotics and Automation (ICRA), 2025, pp. 1–7
[16] I Loncar, J Obradovi c, N Kra sevac, L Mandi c, I Kvasi c, F Ferreira, V Slosi c, D Nad, and N Miskovi c “Marus - a marine robotics simulator,” in OCEANS 2022 Hampton Roads, 2022 pp 1–7
[17] B Lesy, S Herremans, R Kerstens, J Steckel, W Daems, S Mercelis, and A Anwar “Asvsim (airsim for surface vehicles): A high-fidelity simulation framework for autonomous surface vehicle research,” arXiv preprint arXiv:2506.22174, 2025
[18] W Yang, X Wang, X Luo, S Xie, and J Chen “S2s-sim: A benchmark dataset for ship cooperative 3d object detection,” Electronics vol 13 no 5 pp 2024
[19] S Shah, D Dey, C Lovett, and A Kapoor “Airsim: High-fidelity visual and physical simulation for autonomous vehicles,” in Field and Service Robotics, M Hutter and R Siegwart Eds. Cham: Springer International Publishing, 2018 pp 621–635
[20] A H Lang, S Vora, H Caesar, L Zhou, J Yang, and O Beijbom “Pointpillars: Fast encoders for object detection from point clouds,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp 12 697–12 705
[21] T Yin, X Zhou, and P Krahenbuhl “Center-based 3d object detection and tracking,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp 11 784–11 793
[2608.14207] The paper is titled MMUSV-Sim: A Maritime Simulation Platform for Multi-USV Cooperative Perception
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
The MMUSV-Sim platform provides a
Perception-Oriented Maritime Simulation and Data-Generation Platform for Multi-USV Cooperative Perception,
allowing researchers to develop unified workflows for configurable multi-USV scenarios, multimodal acquisition, and shared annotations. -
The system enables the creation of
state-consistent multimodal acquisition
by pausing scene evolution while keeping sensors active, which ensures thatmultimodal observations and world-state records are consistent with one scene state.
-
The platform allows for controllable sensor pose variation through
controllable synthetic heave, roll, and pitch perturbations,
enabling the simulation to varysensor viewpoints and attitudes
based on configurable wave settings. -
The system generates structured exports for downstream tasks by creating
camera-projected 2D boxes, LiDAR-frame 3D boxes, ego-centric BEV boxes, and plan position indicator (PPI) representations.
-
The platform can be used to evaluate cooperative perception performance by testing fusion settings—No Fusion, Late Fusion, and Early Fusion—demonstrating that
Early Fusion improves AP@0.5 from 45.54 to 72.74 relative to the egoonly baseline
under clean conditions. -
The system quantifies geometric agreement between annotations and renderings by computing metrics such as
median center error
andmask containment,
confirming theinternal geometric consistency of the coordinate-transformation and cameraprojection pipeline.
Sources
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation