There is rarely a clear path from a DARPA funding to a commercial product. Defense research typically focuses on finding solutions to particular military issues, resulting in technology that is kept in classified or semi-classified settings for years before anybody outside the original project is aware of its true potential. Similar steps were taken by the autonomous sonar mapping systems that are currently being used in commercial deep-sea exploration: they were first developed for naval surveillance and undersea domain awareness, then gradually declassified, licensed, and modified for the very different economic logic of locating and mapping objects on the ocean floor.
Sparse-aperture sonar is the key technology, and to grasp its significance, let’s take a quick look at the issue it resolves. Conventional sonar mapping uses a single vessel to transmit acoustic energy pulses downward and time how long it takes for the echo to return; the deeper the seafloor, the longer the return time. The sonar sweep, which can be several kilometers wide at deep ocean depths, is covered by a single vessel per pass. The issue is that in order to cover the whole world seafloor in this manner, a vessel must physically transit every location, which would require centuries of continuous operation with the existing fleet of research ships at ordinary survey rates. It becomes clear why only around 28% of the ocean floor is mapped to contemporary standards.
By treating several vehicles as a single, enormous virtual sonar array, sparse-aperture arrays alter that math. The system can achieve the angular resolution of a much larger sonar aperture than any single vessel carries when multiple autonomous surface vessels, or a combination of surface and underwater drones, move in coordinated formation while each one transmits and receives acoustic signals. Higher-resolution mapping over a larger area is the outcome, and researchers are reporting up to fifty times the pace of traditional underwater vehicle surveys. This is made possible by acoustic processing, which combines signals from several spatially separated sources with fractional-second timing precision. Until onboard AI computers became fast enough to perform this computation in real time, it was not practically possible on a moving platform.
What sets this generation of technology apart from previous multi-vehicle sonar approaches is the real-time edge AI component. It takes a lot of computing to process acoustic data from a sparse array, and sending raw data to a shore station or surface ship for processing adds latency and bandwidth restrictions that restrict the system’s speed. Without waiting for human analysis or satellite uplink, the system can map, identify, and highlight seabed characteristics as it passes over them when processing takes place onboard each vehicle on processors tailored for the particular acoustic signal processing task. A potential nodule field, a fault line, a cable, or a ridge feature—all of these are identified in real time and sent as processed results instead of unprocessed sonar data.
As the underlying technology costs decreased and the regulatory framework for operating autonomous maritime vehicles in open ocean conditions became clearer, the commercial shift accelerated. Among the first to use autonomous sonar surveys to monitor cable routes for damage, shifting sediment, and seismic activity that could endanger infrastructure are deep-sea cable operators, the businesses in charge of the submarine internet infrastructure that transports the vast majority of the world’s internet traffic. The monitoring requirements of oil and gas companies with deep-water pipeline infrastructure are comparable, and the cost differential between crewed vessel operations and autonomous surveys is significant enough to make the financial case quite obvious.
The largest potential benefactor of this technology’s broader implementation is the Seabed 2030 program, which seeks to create a comprehensive high-resolution map of the ocean floor worldwide. The number of vessels conducting surveys and the expense of operating them are two factors limiting the project’s current pace, which is adding millions of square kilometers annually. If implemented at scale, autonomous swarms with a coverage rate of fifty times each operational day and a reduced cost per kilometer might drastically shorten the timetable. Insurance for autonomous vessels operating in international waters, data standards for incorporating autonomous survey data into current databases, and the willingness of research funders to invest in fleet deployment rather than individual vessel operations are some of the variables that will determine whether that occurs.
