Trung Tien Dong
ICRA 2027 · Submitted

AquaBEV-Nav: Learned BEV Occupancy for Underwater Navigation and Exploration

Trung Tien Dong1*, Zhenqi Wu1*, Sahasra Kondapalli1, Jiayi Wu3, Yi Sheng2, Xiaomin Lin1

1ERA Lab, University of South Florida  ·  2YES Lab, University of South Florida  ·  3University of Maryland, College Park
*Equal contribution  ·  Also presented as a poster at the WPNMR Workshop @ IROS 2026

arXiv Paper Video Code Coming soon
Pipeline comparison
AquaBEV-Nav. CORAL constructs occupancy through monocular depth estimation, geometric reconstruction, and rasterization (top). AquaBEV-Nav replaces this perception pipeline with direct monocular BEV occupancy prediction (bottom).

Abstract

Safe underwater exploration requires a robot to understand where surrounding structures are located and which regions are available for motion. Existing vision-based underwater exploration systems commonly obtain this information indirectly by estimating monocular depth, unprojecting the geometry into 3D space, and accumulating it into a 2D bird's-eye-view occupancy map. This reliance on intermediate depth estimation is particularly problematic underwater, where scattering and wavelength-dependent attenuation degrade visual cues and limit the reliability of monocular depth estimates.

We introduce AquaBEV-Nav, an underwater exploration framework that bypasses explicit monocular depth estimation through direct bird's-eye-view occupancy prediction. Built upon the CORAL hierarchical exploration framework, AquaBEV-Nav replaces its depth-based perception front end with AquaBEV. Given a single RGB frame, AquaBEV maps visual features into a learned polar representation, performs causal reasoning along the range dimension, and reconstructs local Cartesian occupancy without relying on intermediate depth prediction. The resulting occupancy map is accumulated into CORAL's persistent spatial memory, providing spatial context for VLM-based high-level planning and collision constraints for dynamics-aware local trajectory generation. Across ten simulated reef environments and six occupancy backbones evaluated under a single protocol, AquaBEV-Nav reaches 37.48 structure IoU and 53.2 target IoU, 88.95% closed-loop coverage with zero collisions.

37.48
structure IoU
53.2
target IoU
88.95%
closed-loop coverage
0
collisions across ten simulated reefs

Video

Contributions

Method

AquaBEV-Nav architecture
AquaBEV-Nav architecture. AquaBEV-Nav predicts local structure, target, and observability fields from a single RGB image. These fields are transformed into the world frame and accumulated in the navigation system's persistent spatial memory.

The AquaBEV trunk (image encoder → cross-view transformer to polar features → causal range decoder) feeds separable heads for structure (obstacles) and target (e.g. objects of interest) fields. Training-only objectives — radial transport along each azimuth ray, a distance-weighted false-positive penalty, and target transport — shape these fields for planning without adding parameters. Per-frame predictions are fused into CORAL's persistent map, where a VLM high-level planner proposes waypoints and a dynamics-aware low-level planner generates collision-free trajectories.

Results

Simulation perception
Simulation perception. Qualitative comparison of local BEV predictions across representative environments.
Closed-loop trajectories
Closed-loop navigation by reef topology. Executed trajectories produced by different perception front ends under the same CORAL navigation stack.
Pool perception
Pool perception. Front-view RGB and local BEV predictions from AquaBEV-Nav and the depth-based CORAL front end.
Real-world qualitative results
Real-world transfer. Pool and Blue Grotto field frames: AquaBEV-Nav vs. CORAL and GaussianFormer.

BibTeX

@misc{dong2026aquabevnav,
  title         = {AquaBEV-Nav: Learned BEV Occupancy for Underwater Navigation and Exploration},
  author        = {Trung Tien Dong and Zhenqi Wu and Sahasra Kondapalli and Jiayi Wu and Yi Sheng and Xiaomin Lin},
  year          = {2026},
  eprint        = {2609.32156},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url           = {https://arxiv.org/abs/2609.32156}
}