umfieldrobotics/OceanSim

[IROS 2025] OceanSim: A GPU-Accelerated Underwater Robot Perception Simulation Framework

What it solves

OceanSim addresses the difficulty of developing robust underwater perception systems by providing a high-fidelity simulation environment. It aims to reduce the "simulation-to-real gap" by accurately modeling how visual and acoustic sensors behave in underwater environments.

How it works

Built on NVIDIA Isaac Sim and the NVIDIA Omniverse ecosystem, the framework uses GPU-accelerated parallel computing and physics-based rendering to simulate underwater sensor data. It utilizes OpenUSD for efficient 3D workflows and integrates with ROS 2 for sensor publishing and robot control.

Who it’s for

It is designed for researchers and developers working on underwater robotics, specifically those focusing on perception solutions and the generation of synthetic data for training underwater object detection models.

Highlights

  • GPU-accelerated rendering via NVIDIA Isaac Sim.
  • Physics-based simulation of both visual and acoustic sensors.
  • Native ROS 2 integration for control and sensor data.
  • Support for OpenUSD 3D workflows.
  • Open-source framework for community-driven development.

Related

  • Project
    isaac-sim/IsaacLab

    A GPU-accelerated open-source framework built on NVIDIA Isaac Sim for simplifying robotics research in reinforcement learning, imitation learning, and motion planning.

  • Project
    mani-skill/ManiSkill

    An open-source robot simulation and training framework that uses GPU parallelization to enable high-throughput data collection and training for manipulation skills.

  • Project
    Cosys-Lab/Cosys-AirSim

    A high-fidelity simulator for drones and cars built on Unreal Engine, providing advanced sensor simulation and ground-truth data generation for autonomous vehicle research.

  • Project
    abizovnuralem/go2_omniverse

    A high-fidelity digital twin project for Unitree Go2 and G1 robots in Nvidia Isaac Sim, enabling real-time synchronization between physical robots and their virtual counterparts.

  • Project
    NVlabs/ProtoMotions

    A GPU-accelerated simulation and learning framework for training physically simulated digital humans and humanoid robots, supporting large-scale motion learning and sim-to-real deployment.