A computationally-efficient 2D imaging sonar model for underwater robotics simulations in Gazebo

Abstract

Divers have to perform technical underwater tasks in dangerous and unstructured environments. To reduce a diver's workload and improve overall safety, an underwater robotic assistant (UWRA) could be deployed with the diver. The UWRA could assist the diver in navigation, quickly ferry tools from the surface, and carry underwater samples. To develop and test the autonomy required for such a UWRA, autonomy developers could benefit from a simulated underwater environment. Unfortunately, a real-time simulation for one of the most important sensors in underwater robotics, a forward-looking 2D imaging sonar, does not exist in robotics simulators. We developed a simulation for 2D imaging sonars that executes in real-time in the Gazebo robotics simulator. The 2D imaging sonar simulator was assessed by comparing the sonar images that it generated with sonar images generated by real sonar systems in similar environments. For example, we tested both the simulated and real sonar systems in environments with underwater man-made structures, divers, and varying terrain. The sonar simulator was also assessed based on its computational complexity and compared to other similar simulators with respect to sonar image generation rates. We found that our 2D imaging sonar simulator generated qualitatively realistic images and that it was computationally efficient enough to execute in real-time with the Gazebo simulator.