A Multi-Beam Forward-Looking Sonar Simulator Based on Unity3D

The development and testing of robust underwater acoustic technologies for applications such as environmental monitoring, inspection of underwater structures, as well as applications in the fields of geology and marine biology, typically require large amounts of data, which involves significant effort and costly data collection campaigns at-sea due to the challenges posed by underwater environment. Consequently, a realistic sonar simulator capable of generating realistic synthetic acoustic data at a significantly lower cost offers a viable alternative to accelerate and enhance the implementation of such algorithms. This work proposes a Multi-beam Forward-Looking Sonar (MFLS) simulator developed within the Unity3D simulation environment. The simulated sensor is capable of producing synthetic datasets that are valuable for training and optimising perception algorithms, and it can be easily integrated in Unity-based robot simulators. The MFLS model proposed in this paper employs ray-tracing techniques exploiting the RayCast functionality already available in Unity3D, and includes the modeling of transmission loss, acoustic backscattering, and speckle noise phenomena. Preliminary results obtained by comparing real acoustic images with simulated ones demonstrate that the simulator provides data that are qualitatively consistent with real acoustic images. Future developments include further investigation of the already implemented acoustic propagation phenomena and the development of new ones, integrating the simulated sensor into the Robot Operating System (ROS) framework, and ultimately validating results by training neural networks on synthetic data generated by our sensor model and testing them with real acoustic images.