HoloOcean 2.0: Next-Gen AUV Simulation

• HoloOcean 2.0 is a next-generation simulation platform for AUVs that integrates physically validated six-DOF dynamics, ROS 2 interfaces, and realistic sensor modeling.
• It employs a modular design with Unreal Engine 5.3 visualization and a Python-based dynamics manager, enabling both SIL and HIL workflows for comprehensive testing.
• Performance validation shows simulation metrics closely matching real-world trials, thus accelerating autonomous underwater research and system tuning.

HoloOcean 2.0 defines a new generation of high-fidelity marine robotics simulation for autonomous underwater vehicles (AUVs), integrating physically-validated hydrodynamics, cutting-edge graphical realism, and ROS 2–native interfaces. It addresses longstanding challenges in scalable, closed-loop testing of AUV navigation, control, perception, and autonomy algorithms by enabling hardware-in-the-loop (HIL) and software-in-the-loop (SIL) workflows on standard laboratory hardware, while providing extensibility for sensor, vehicle, and environment modeling (Meyers et al., 10 Nov 2025, Romrell et al., 7 Oct 2025).

1. System Architecture and Modular Design

HoloOcean 2.0 represents a multi-tier architecture combining Unreal Engine 5.3 for visualization and collision with an external, modular dynamics manager based on Fossen’s six-degree-of-freedom marine vehicle models (Romrell et al., 7 Oct 2025, Meyers et al., 10 Nov 2025). Vehicle dynamics are managed in Python, orchestrated by a per-tick “dynamics manager” that computes state updates and exchanges forces with Unreal’s physics system. This conceptual decoupling permits drop-in replacement or augmentation of vehicle models and direct user control over hydrodynamic, hydrostatic, and control-surface parameters.

Major modules include:

• Vehicle Dynamics: Python-implemented Fossen 6-DOF dynamics, supporting user-specified parameters via YAML or Python interfaces.
• Sensor Simulation: Cameras (RGB, depth, semantic), ray-tracing–based sonars (echo, sidescan, bathymetric), and LiDAR.
• Environment Management: Static and procedurally generated landscapes, volumetric effects (currents, turbidity), and realistic water rendering.
• ROS 2 Integration: A native Python/C++ ROS 2 bridge for synchronized data streaming and control, using both standard and custom message types.
• Visualization Pipeline: Features Lumen dynamic GI, Nanite virtualized geometry, realistic water shader plugins, and support for spectral/Gerstner wave models.

This design enables scalable simulation: users can configure and instantiate a variety of vehicles, sensor payloads, and environmental conditions, rapidly transitioning between SIL and HIL modalities (Meyers et al., 10 Nov 2025).

2. Underwater Vehicle Dynamics and Control

At the core of HoloOcean 2.0, vehicle motion is governed by a full six-degree-of-freedom, Fossen-inspired model:

$$M\dot{\nu} + C(\nu)\nu + D(\nu)\nu + g(\eta) = \tau, \quad \dot{\eta} = J(\eta)\nu$$

Here, $\nu$ denotes body-fixed velocities, $\eta$ denotes inertial position and attitude, $M = M_{RB} + M_A$ is total inertia, $C(\nu)$ is the combined Coriolis/centripetal matrix (rigid-body + added mass), $D(\nu)$ is nonlinear hydrodynamic damping, $g(\eta)$ restores gravity/buoyancy, $J(\eta)$ maps velocities to inertial derivatives, and $\tau$ aggregates external forces and moments (thrusters, fins) (Romrell et al., 7 Oct 2025, Meyers et al., 10 Nov 2025).

Torpedo fin forces are computed as:

$$f_i = \frac{1}{2}\rho A_i C_{L,i} v_{r,i}^2 \delta_i$$

$M_i = R_i \times F_i$

where $R_i$ is the fin’s offset, $v_{r,i}$ is the local relative speed, $A_i$ and $C_{L,i}$ are fin area and lift coefficient, and $\delta_i$ is actuator input. All contributing forces and moments are summed into $\tau$. Control surfaces employ first-order servo models with user-tunable time constants for realistic actuator response.

User interaction is facilitated through YAML and GUI-based configuration for vehicle mass properties, hydrodynamic coefficients, fin geometry, and time constants. This framework supports direct parameter tuning for both model-based control and empirical identification workflows.

3. Sensor and Environment Simulation

Sensor models leverage Unreal Engine’s built-in and custom ray-tracing APIs. For sonar, each simulated beam uses the UE5 API for per-tick world intersection:

• Define beam set $\{d_i\}$ for field of view (FOV).
• Each tick: for each $i$, calculate hit location via $World.Raycast(origin, d_i)$.
• Compute range/intensity using standard sonar equations.

Performance data indicates that ray tracing yields a $\sim4\times$ speedup relative to octree-based methods (Romrell et al., 7 Oct 2025). Sensor data is formatted as standard ROS 2 messages ($sensor\_msgs/Image$, $PointCloud2$, custom $holoocean\_msgs/SonarPing$). Semantic sensors annotate per-pixel or per-point object class labels as required.

Environmental features include procedural terrain, bathymetry import (GeoTIFF→Unreal), and volumetric actors for currents or turbidity, settable via user parameters or procedural Blueprints. Water surface models include both Gerstner wave superpositions and spectral FFT-based approaches, enabling future hydro-physical coupling.

Current limitations include the lack of world-level current/wave physics in Fossen’s model and non-implementation of full sonar semantic reflection and multibounce (Romrell et al., 7 Oct 2025).

4. ROS 2 and HIL/SIL Workflows

HoloOcean 2.0’s ROS 2 integration is provided via a bidirectional bridge:

• holoocean_simulator: Unreal-side node, handling sensor publication and control message subscription.
• holoocean_client: Python node, mirroring hardware driver API and relaying ROS topics/services/actions.

The system employs standard message types:

Sim Topic	ROS 2 Type / Custom
/holoocean/imu	sensor_msgs/Imu
/holoocean/pressure	sensor_msgs/FluidPressure
/holoocean/dvl	sensor_msgs/Range, nav_msgs/Odometry
/holoocean/depth	std_msgs/Float32
/holoocean/actuators	holoocean_msgs/ActuatorSetpoint
/holoocean/control	geometry_msgs/Twist

The bridge supports both HIL (with real AUV onboard software/hardware) and SIL (entirely simulated control stack). This framework allows controller software, such as for the CougUV torpedo AUV, to be tested without modification across lab and simulated environments (Meyers et al., 10 Nov 2025). Sensor models include tunable noise (Gaussian bias/drift for IMU, noise for pressure sensors, etc.), configurable dropouts (for DVL), and user-defined latency profiles.

5. Validation and Performance Results

Validation using canonical vehicle behaviors includes depth-hold, heading-hold, and altitude-hold (relative to the seabed), comparing simulated results against real underwater trials (Meyers et al., 10 Nov 2025):

• Depth-Hold: Step 1m→3m input. Simulated rise time ≈ 8s, overshoot <5%; real-world ≈ 9s, ≈7% overshoot, RMSE ≈ 0.08m.
• Heading-Hold: 30° yaw step, sim → ±2° band in 12s; real 13s underwater. Surface-level degradation in real due to unmodeled fin effects.
• Altitude-Hold: On flat/bumpy terrain, sim error σ ≈ 0.1m; real σ ≈ 0.12m on gentle slopes.

Mission Type	Avg. CPU (%)	Avg. RAM (%)
Real-World	27.4	8.3
HIL Simulation	26.5	8.2

HIL/SIL simulations execute in real time on commodity Intel i7 hardware, with negligible computational overhead from the custom dynamics or ROS 2 bridge layers (Meyers et al., 10 Nov 2025).

6. Comparative Analysis and Limitations

Distinctive advancements over HoloOcean 1.x include migration to true modular Fossen dynamics (decoupled from Unreal’s engine), native ROS 2 bridge (abandoning third-party wrappers), Dockerized HIL/SIL workflows, and real-time performance on embedded-class hardware (Romrell et al., 7 Oct 2025, Meyers et al., 10 Nov 2025). Sonar simulation, semantic sensor support, and high-lod environment rendering have also been substantially upgraded.

Current limitations:

• No integration of world-level currents or wave–surface dynamics into physical models (notable in heading-hold near surface).
• Acoustic ranging and multi-vehicle acoustic communication are in development.
• System-identification tools for automated parameter refinement are not yet integrated.
• Only torpedo-style AUVs are modeled; surface and hovering vehicles are pending support.

7. Application Scope and Future Directions

HoloOcean 2.0 provides a validated, open-source platform for:

• Accelerated AUV controller development and tuning with direct transfer from simulation to field deployment with minor retuning.
• Systematic evaluation of autonomy, perception, and SLAM algorithms—including for large-scale, out-of-lab scenarios.
• Sensor and acoustic system development in virtualized, highly configurable marine environments.

Ongoing and future work cited includes expansion of multispectral sensor simulation, integration of on-the-fly system identification, automated scenario/environment generation, and support for high-resolution acoustic communications and species-level environmental perception (Romrell et al., 7 Oct 2025, Meyers et al., 10 Nov 2025, Mallery et al., 2019).

References: (Meyers et al., 10 Nov 2025, Romrell et al., 7 Oct 2025, Mallery et al., 2019)