Uncrewed Systems Technology 049 - April/May 2023

102 often around 100-200 mm in length and diameter, and consume about 20-30 W, to scan with 1o horizontal beamwidths at frequencies of 400-900 kHz. Probably the smallest in the world at the moment weighs about 0.3 kg and consumes up to 240 mA at 24 V, and can operate at depths of 6000 m. By contrast, the new breed of 3000- plus kg XLUUVs being built for upwards of 2000 km range might incorporate forward-looking sonars that are more than a metre in length and emit pulses at ultra-low frequencies in order to see objects almost a kilometre ahead. While higher frequencies would yield sharper images, long-range detection is critical for defence missions such as subsea border monitoring and asset protection, as well as for surveys in polar regions where icebergs or ice shelfs are unmapped and so represent critical safety hazards to vessels. Some particularly advanced forward-looking sonars can even use a combination of interferometric and Vernier processing to map out the seabed terrain ahead of the UUV as it moves parallel with the seafloor. Real-time data outputs include vertical height discrimination of targets ahead of the sonar. Such systems incorporate sub- arrays in configurations that allow switching between a 3D mode to enable bathymetric terrain mapping ahead of the uncrewed vehicle, and a 2D high- resolution imaging mode that can acoustically inspect features of interest in greater detail. At the time of writing, military users were integrating this technology into their UUVs in order to perform simultaneous forward-looking obstacle avoidance and forward-looking terrain mapping, and the technology could soon be used to power high-level autonomous behaviours such as underwater SLAM. A key area of further development of forward-looking sonars in the near future will be the application of ML for real-time object detection and classification, as is being done with some bathymetric sonars, for instance to better recognise objects such as mines, minerals or potentially hostile UUVs. Autonomous sonar survey missions could soon be regularly gathering thousands of terabytes of data per day though, which represents something of a bottleneck in terms of how best to process it in order to support decision-making by their operators. ML will therefore be critical for designing computer-aided detection April/May 2023 | Uncrewed Systems Technology The use of AI techniques is enabling cleaner sonar data as well as better object detection, following, and analysis (Courtesy of Teledyne) Focus | Sonar systems Historically, transducers have been made using difficult materials and processes, but these and their costs have eased in recent years (Courtesy of Blue Robotics)