Publications detail
FinsSim: A Reality-Aligned Integrated Simulation Platform for Underwater Robot Learning
Abstract
Underwater robot learning relies on simulators that integrate high-fidelity hydrodynamics, convenient learning interfaces, and a credible transition to real scenarios. In this work, we present FinsSim, a reality-aligned integrated simulation platform for Sim-to-Real underwater robot learning. FinsSim first constructs high-fidelity simulation with selectable backends to adapt to diverse requirements. To facilitate underwater robot research, it further offers standard control baselines, alongside with unified robot learning workflows. For reliable Sim-to-Real transfer, FinsSim adopts a multi-sensor fusion scheme to provide low-cost yet precise localization. Moreover, it implements calibrated thruster-hydrodynamics models and a constrained wrench allocation algorithm. Bridging these modules by ROS 2, FinsSim establishes a complete Sim-to-Real transfer pipeline. Through matched simulations and experiments, it is demonstrated that reliable Sim-to-Real transfer of underwater robot control policies can be achieved with the FinsSim framework. Separate ablation studies also validate that the modules of FinsSim can address the pivotal issues of underwater Sim-to-Real from different aspects. Overall, this work aims to bridge the gap between theoretical research and practical applications, ultimately driving advancements in the field of underwater robotics.
Yu Zhang, Yuanmingqing Song, Xiangyun Rao, Pangkit Fong, Kunhao Zhang, Chongrong Fang, and Jianping He. “FinsSim: A Reality-Aligned Integrated Simulation Platform for Underwater Robot Learning.” arXiv:2609.23943, 2026.
01
High-fidelity simulation
FinsSim supports selectable simulation backends for different modeling, training, and evaluation requirements. Its interchangeable hydrodynamic models span Simplified Fossen, Calibrated Fossen, and Physics-Mesh paths.
This makes the trade-off between computational cost, modeling convenience, and physical alignment explicit instead of fixing research to a single simulator configuration.
02
Unified learning workflows
The platform provides standard control baselines, single- and multi-agent learning interfaces, and parallel-training workflows. Sensor simulation and unified observation and control interfaces keep data collection, policy training, simulation control, and evaluation in one frame.
These workflows support common underwater-control research tasks without binding the platform to one algorithm or vehicle.
03
Reality-aligned deployment
FinsSim combines low-cost multi-sensor fusion with calibration of thrusters, actuators, and hydrodynamics to align the simulation model with the physical platform. Constrained wrench allocation maps policy or controller commands safely to physical thrusters.
ROS 2 bridges simulation, localization, control, and hardware communication into a reusable simulation-to-hardware pipeline.
04
Experimental verification
Matched simulations and pool experiments evaluate localization accuracy, hydrodynamic replay, four-degree-of-freedom station keeping, and three-dimensional trajectory tracking. Ablations isolate the role of the key modules.
The Calibrated Fossen reference policy sustains 7 of 8 station-keeping setpoints. Long-horizon tracking covers straight, broken-line, elliptical, and lemniscate references; the project page presents the supporting figures and video.