A flotilla of small boats swarmed into the Solent in a large-scale MOD exercise to aid in the develop Artificial Intelligent (AI) systems in identifying craft and their occupants.
A total of 13 vessels mimicking close formations, convoys, attack profiles and beach landings were conducted, supported by some 130 personnel, multiple drones, a light aircraft and more than 50 cameras and sensors were involved in the five-day exercise, spread out across 18 miles of the Solent.
Run by experts from the Defence Science and Technology Laboratory (Dstl) – who harness leading-edge science and tech to benefit the UK’s Armed Forces – and the Royal Navy, the goal was to gather a mass of data (stills and video) of craft moving around on the water before coming ashore.
The footage and stills, plus other data – visual, infrared, sonar and radar sensors were all used – recorded during the exercise will now be fed into software under development, as well the relevant ‘metadata’ including platform and sensor locations, weather, sea states and other contextual information.
The algorithm needs what experts call a ‘training dataset’ to help AI to identify both craft and the people aboard, determining whether they are acting in a co-ordinated, controlled manner (e.g. military personnel), or haphazard, random way (civilians).
The exercise extended from Hurst Castle in the west to Browndown near Gosport. Contending with high winds up to 40 knots, landings saw personnel boarding and leaving craft in different ways to generate data representative of different behavioural traits.
In one scenario synchronised beach landings saw participants acting as a trained team – like a military unit. In another, participants left boats in a deliberately chaotic way, to provide a wider data sample of human movement.
The trial was a collaboration with 12 industry and international partners, supported by British sailors and soldiers, and took place ahead of the UK government’s AI Summit.
“This was an ambitious and challenging trial which builds on the experience and expertise gained during a previous land-based exercise,” said Charlie Maslen, Dstl’s trial technical authority.
“Conducting a trial with sensors spanning three domains – land, sea and air – involving 12 separate industry partners was immensely complex, added to which we were hampered on two days by 40-knot winds.”
Data generated by the trial will also enable the Ministry of Defence and industry partners to develop new AI products across defence in an ethical, safe, and responsible manner.
