A Bayesian Occupancy Grid Mapping Method for the Control of Passive Sonar Robotics Surveillance Networks

We address the problem of building the perception layer for controlling a network of autonomous, sensorised robots in an underwater surveillance application. To this purpose, we propose a novel Occupancy Grid (OG) mapping framework, based on an adaptation to the standard OG method. The method iteratively builds an OG map from successive acoustic measurements. The resulting maps, produced in real-time by each robot, show the probability of each grid cell containing a target. The algorithm is designed to handle the presence of multiple targets and takes into account a dynamic world (moving targets). Albeit generic, the proposed method was adapted to work with passive sonar sensors that produce bearing-only measurements. Results from simulations demonstrate the capability of the method to create maps showing regions likely containing targets. They also show how data fusion between maps produced with sensors spatially separated can achieve target localisation and tracking. The produced OG maps provide valuable guidance for robot cooperative decision making. They identify interesting areas to survey (higher target presence probability), regions likely empty and areas not adequately monitored.