Real-time sonar simulation for UUVs using rasterization and omnidirectional screen space reflections

Since real-world testing can be costly, risky, and difficult to reproduce, sonar simulation has become an important tool for developing and validating perception and navigation algorithms for unmanned underwater vehicles (UUVs). This work introduces a real-time sonar simulation framework integrated into the NauSim platform. Based on GPU rasterization and an omnidirectional extension of screen space reflections (SSR), the framework efficiently simulates primary echoes and approximates multipath effects. The proposed framework incorporates several acoustic and imaging phenomena, including Lambertian backscattering, signal attenuation, speckle noise, acoustic shadowing, and the influence of side lobes through beam-pattern modeling, while maintaining the computational performance required for real-time operation. In addition, dedicated anti-aliasing techniques, including procedural dithering and microtexture perturbations, reduce artifacts inherent to rasterization-based approaches. The framework is validated using real sonar data acquired with a Ping360 sensor mounted on a BlueROV2 platform. Quantitative and qualitative evaluations show good structural agreement between simulated and real sonar images. These results demonstrate that the framework provides a practical, computationally efficient tool for generating synthetic sonar datasets, thereby supporting the development and evaluation of autonomous underwater perception algorithms.