A Virtual Testbed for the Development of Multi-sensor Perception Systems in ROV-Based Underwater Operations

The development of multi-sensor perception systems is essential to support advanced underwater robotic operations such as inspection, mapping, and manipulation, typically carried out with Remotely Operated Vehicles (ROVs). However, field testing is often costly, logistically demanding, and constrained by safety and accessibility issues. This paper presents a flexible virtual testbed for the development and validation of perception pipelines in ROV-based missions. The framework integrates Unity game engine, for high-fidelity 3D simulation, with Robot Operating System 2 (ROS2), for real-time sensor processing, robotic control, and data exchange. The testbed provides a modular emulation of ROV architectures, integrating digital vehicle replicas, configurable underwater scenes, and simulated sensors such as stereo vision, forward-looking sonar, and inertial units. This setup enables the generation of synthetic but physically consistent data streams, including depth maps, acoustic imagery, and fused spatial reconstructions. An operator console developed in Unity integrates multiple Augmented Reality (AR) strategies to enhance situational awareness and manipulation accuracy. The proposed virtual testbed offers a cost-effective and flexible environment for iterative development, debugging, and operator training. It provides a valuable tool for accelerating the design and integration of perception systems in underwater robotics, especially in scenarios where field trials are impractical or high risk.