Eavesdropping like a bat: Towards fusing active and passive sonar for a case study in simultaneous localization and mapping

Abstract

Among so-called active sensors that use self-generated signals, sonar sensors are more challenging to implement than lidar and radar due in part to their limited angular field of sensing. A common solution to this challenge is scanning sensors that sweep an angular range with successive measurements. However, scanning sensors are particularly problematic for sonar because of the relatively slow sound speed and the inertia of the sonar head. Studies of bat behaviour suggest that bats may eavesdrop on their conspecifics during group flight. In other words, they fuse information gathered by their own active sonar with information they receive by passively listening to peers. Because bats are extremely skilled in using sonar, this behaviour inspired an investigation into whether fusing active and passive sonar can be a solution to the challenges of implementing sonar sensors. A model of fused sensing is defined, and a numerical simulation is used to answer this question on the test bed problem of simultaneous localization and mapping (SLAM). The simulation results show that when the angular range of active sonar and associated noise is relatively small, the robot's performance in solving SLAM is improved.