Monday, 26 October 2026 · Taipei
TaiwanDefenceNews · Analysis · Intelligence
Latest

Taiwan's maritime council says an AI coast guard drone swarm has passed a 20-aircraft simulation

A council-funded university team has flown two simulated swarms of mixed vertical-take-off drones together and now plans hardware tests and flights at sea.

T
TaiwanDefence

Taipei ·

2 min read
A small tail-sitting ducted-fan drone flying over the sea beside a grey US Navy amphibious ship
File photo: a V-BAT drone, one of the three aircraft types modelled in the Taiwanese swarm project, flies near the US amphibious ship USS Rushmore in the Philippine Sea in August 2022. The photo is not from the Taiwanese research.U.S. Marine Corps photo by Lance Cpl. Manuel Alvarado (public domain) public_domain_verified

TAIPEI —

Taiwan's Ocean Affairs Council, which oversees the Coast Guard Administration, said on Monday, 26 October, that a research project it funds has completed a synchronised flight simulation of two drone swarms totalling 20 aircraft. The council presented the result as a step toward patrols that combine ships with groups of unmanned aircraft working together, the Liberty Times, the China Times and the defence ministry's Youth Daily News reported.

The work is being carried out by a team from the Department of Aeronautics and Astronautics at National Cheng Kung University in Tainan. The project covers artificial intelligence mission planning for mixed fleets of fixed-wing drones that can take off and land vertically, and it is meant to support what the council calls a three-dimensional coast guard, in which surface patrols are backed by aircraft overhead.

According to the council, the team modelled three commercially available vertical take-off designs: the JUMP 20, the V-BAT and the Flexrotor. It built three-dimensional models of each airframe, analysed their aerodynamic parameters and ran six-degree-of-freedom flight simulations, producing flight-behaviour and control models that let the system assign tasks according to the strengths of each type.

The control software relies on multi-agent reinforcement learning, a form of machine learning in which several agents learn to act together. The council said this allows the drones to form up much like a flock of geese, sharing out tasks, planning routes and coordinating with one another on their own. The team added leader-follower formation control and dynamic obstacle avoidance so the swarm can change course when conditions shift or obstacles appear.

To prevent collisions within the swarm, the algorithm enforces three-dimensional spacing between aircraft and a minimum vertical separation of 100 feet, or about 30 metres. The council described the simulation as a shift from controlling single aircraft to coordinating several fleets at once.

The next stages are hardware-in-the-loop testing, integration of prototype aircraft and flight trials over the sea, which the council said will check whether the system works reliably in different sea conditions. The reports did not give a timetable, a budget, or say whether the coast guard intends to buy any of the three aircraft types.

The council linked the project to the pressures on its patrol fleet. It said traditional sea patrols cover wide areas in changeable weather and depend heavily on crews and ships, and that law enforcement needs and grey-zone harassment are driving its push for air cover. It also said recent regional conflicts show drone use moving from single aircraft to coordinated swarms of mixed types. In the council's view, the system could act as the eyes of coast guard crews, improve maritime defence and rescue work, and strengthen whole-of-society defence resilience.

Sources

Have information about this story? Read how to contact the newsroom →

Discussion (0)

Comments are not currently available.

  • No approved comments yet.

More in Industry