116 Underwater Drones Earn Guinness Record for Largest Swarm
Guinness World Records has recognized a trial of 116 TH-618 underwater drones as the largest underwater drone swarm in history, according to Engadget.
The trial took place in August 2026 in a pool measuring 164 feet by 65 feet. Researchers at Northwestern Polytechnical University ran it with Xi'an Tianhe Maritime Technologies, a subsidiary of the defense contractor Xian Tianhe Defense Technology.
Each TH-618 is roughly 2 feet long, half a foot wide and weighs 11 pounds. It carries four propellers and a camera, and is rated for a range of just over 18 miles and roughly 8 hours of continuous operation. For navigation, each unit uses smartphone cameras mounted on its front.
During the run, the drones executed formations and maneuvers autonomously. Over roughly 34 minutes, they relied solely on visual data to move around the pool, covering more than 1,600 feet without human intervention.
The navigation method is the detail that stands out. GNSS, the satellite positioning family that includes GPS, does not penetrate water, so most underwater operators rely on acoustic positioning, which uses sound to locate vehicles, or on tethered control, where a cable links the vehicle to an operator. A swarm holding formation on onboard cameras alone has to solve visual odometry (estimating motion by tracking how the scene shifts from frame to frame), obstacle avoidance and inter-vehicle spacing. It does all of this on limited compute per node, while water clarity, lighting and turbulence vary.
In my view, the scale of the number tells less than the conditions around it. A 164 by 65 foot pool has known geometry, still water and no meaningful currents. Open-water deployments would introduce variables a pool does not, and the gap between a successful pool trial and a field-ready system is where much of the engineering effort usually sits. That caveat does not diminish the result. It places the result accurately.
The affiliation is worth stating plainly. Xi'an Tianhe Maritime Technologies is a subsidiary of a defense contractor, and the trial was run with a university partner. Underwater swarms have obvious relevance to maritime surveillance, mine countermeasures and seabed inspection.
For engineers, the most useful takeaway is that 116 vehicles completed a 34-minute autonomous run on cameras alone, in a medium where that is difficult. Whether the approach transfers to turbid or open water, and at what compute and power budget, is what the next round of published data would need to answer. If vision-based navigation holds up beyond the pool, it could cut sensor cost and weight per vehicle, which matters for any swarm aiming to grow beyond a few dozen units. For now, this is a documented demonstration in a constrained setting, and it is worth reading as one.
