appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Stephen Chen

A research team from northwestern China has released a new algorithm that could fundamentally change how drone swarms hunt and destroy enemy targets.
The algorithm, known as HG-STR (Heterogeneous Graph Spatio-Temporal Reasoning), promises to allow a fleet of fixed-wing drones to autonomously search a vast battlefield and eliminate every single enemy, even when their communications are being jammed and their vision is blocked.
In a simplified, ideal environment, the algorithm could achieve a 100 per cent coverage rate, an increase on other algorithms, while operating fast enough to keep up with the pace of modern warfare, according to a peer-reviewed paper published in China’s top aviation journal Acta Aeronautica et Astronautica Sinica on May 19.
Most drone operations today are still controlled remotely by human pilots, according to a Beijing-based defence expert who was not involved in the study.
“This technology suggests a future where swarms of drones could be sent into a high-risk, jammed environment, cut off from human command with a single final order: find and kill them all,” he said, requesting anonymity because he is not authorised to speak to media.
Traditional algorithms treat all information, such as friend, foe and terrain, as the same type of data.
This creates confusion, according to the paper.
The new method builds a “heterogeneous graph”, a kind of smart web where every object is tagged with its true meaning. A friendly drone is one type of node, a search area is another and an enemy target is a distinct third.
The algorithm can learn to pay attention to the right connections. When a drone spots a target, that information is treated as a high-priority threat. When a teammate is nearby, it is treated as a collaboration opportunity.
This allows the swarm to instantly understand whom to help and whom to hunt, according to the paper.


Older rule-based systems, like a pre-written script, fail when the enemy does not follow the script. Most existing optimisation methods, like a chess computer calculating every move, are far too slow.
In the heat of battle, they take seconds to decide and, “in this period, a drone can fly nearly 600 metres (1,968 feet) blind, a fatal delay in intense electromagnetic confrontations”, the paper stated.
The HG-STR algorithm makes decisions in just 6.6 milliseconds, a huge leap over older methods.
But the real battlefield is full of chaos. Enemy jamming can cut communication links, leaving drones isolated. A drone’s own sensors can only see a small patch of ground. The “search and destroy” mission is a desperate race against time and fuel.
The Chinese team faced these challenges head-on. First, they gave each drone a “memory”. When a drone loses contact with its teammates, it relies on a special memory module (a GRU, or Gated Recurrent Unit) to remember where its friends were and where the enemies were last seen.
Second, they designed a hierarchical brain. The drone first decides its goal: “Should I search or should I strike?” Then, it chooses a specific target. Finally, it decides how much ammunition to use.
By breaking the problem into layers, the drone avoids the chaos of trying to do everything at once.
In the team’s simulation experiments, even when the communication radius was severely limited, the algorithm achieved a 100 per cent kill rate of enemy targets, including those hidden from plain sight.
The experiments also showed the algorithm can be trained on a small scenario and then instantly deployed to a much larger one, with more drones and more targets, without retraining.
Looking to the future, the team plans to take the algorithm from the lab to the real world. They also want to make it even more robust, teaching it to handle not just lost connections, but also delayed messages and corrupted data.
“Future work will focus on exploring the lightweight deployment and real-flight verification of this algorithm on embedded airborne platforms with limited computing power,” the paper stated.


China tested a handheld submarine detector. It tracked a subway train underground

How Wang Zhenyi, Chinese female astronomer and mathematician, defied gender expectations

Chinese drug loberamisal extends stroke treatment window from 6 hours to 2 days

China-Russia space camp in Malaysia opens to youth in Southeast Asia

Science

Chinese scientists create self-deciding algorithm for drone swarm

It is the first known algorithm capable of achieving a 100 per cent kill rate while operating fast enough to keep up with the pace of modern warfare, a recently published paper says. Photo: Shutterstock

Why tiny atomic clocks may hold the key to China mass-producing cheap swarm drones

The team’s atomic clock, which is said to lose only a second every 30,000 years. Photo: Handout

Landmark in-air refuelling test extends China’s drone swarm delivery range to US homeland

This image shows a drone-to-drone refuel test flight with two signalling bulbs being blocked. Photo: NPU

Will string of science scandals ruin century-old journal Nature’s reputation in China?

A former biomedical PhD student turned whistle-blower and science blogger, known as “Student Geng”, has made a string of claims about the integrity of articles in the prestigious science journal. Photo: Bilibili

Tongji University in Shanghai has taken disciplinary action against a top cancer researcher. Photo: Handout

Omega-3 could hurt the brain, China scores rare-earth find: 7 science highlights

A rare earth mine at Bayan Obo, in the Inner Mongolia autonomous region of China, seen in 2011. Photo: VCG via Getty Images

Many older adults are accustomed to taking omega-3 supplements to help protect cognitive function. Photo: Shutterstock

A worker poses at a rare earth processing plant not far from Bayan Obo. Photo: Sinopix

Many of the C919’s most critical components are supplied by Western firms. Photo: Dickson Lee

The J-20 or “Mighty Dragon” is seen as China’s answer to the American F-22 Raptor. Photo: Captured from 81.com

An apple orchard in China’s western Xinjiang Uygur autonomous region, along the edge of the Taklamakan Desert. Photo: Xinhua

How China and Russia could team up to shift global science ‘centre of gravity’

Russian President Vladimir Putin (right) and Chinese President Xi Jinping (left) walk together during their meeting in Beijing on May 20, 2026.  Photo: EPA/ Sputnik/ Kremlin Pool

News

China

Smart web identifies objects as friend, foe or terrain, meaning drone swarms can hunt even when communication is jammed and vision blocked


Smart web identifies objects as friend, foe or terrain, meaning drone swarms can hunt even when communication is jammed and vision blocked.


National Natural Science Foundation of China

Northwestern Polytechnical University (China)

Chinese scientists create self-deciding algorithm for drone swarm

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Smart web identifies objects as friend, foe or terrain, meaning drone swarms can hunt even when communication is jammed and vision blocked

Smart web identifies objects as friend, foe or terrain, meaning drone swarms can hunt even when communication is jammed and vision blocked