appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Chao Kong

A Chinese coastguard ship is shielding a cluster of fishing boats when several small, fast vessels begin circling – sometimes darting closer, sometimes backing off. Are they just curious fishermen? Or are they probing for a weak spot, trying to lure the patrol ship into a trap?
In the real world, a split-second decision could escalate a tense maritime stand-off into something far more dangerous. This is exactly the kind of nerve‑wracking scenario Chinese coastguard researchers have been running inside a computer war game.
Their findings, published this month, revealed a surprising number: under the existing rule‑based command system, the simulated Chinese coastguard would resort to lethal weapons only 0.3 per cent of the time.
It is a remarkably low risk – but for Beijing, even that chance is too high when the encounters involve sensitive, disputed waters.
To push the risk down to zero, the team developed an artificial intelligence (AI) framework that can read a vessel’s intentions like a seasoned mariner reads the sea.
The AI does not simply stamp a target as “hostile” or “harmless”; it continuously calculates the odds of multiple possible intentions – from routine fishing to coordinated harassment – and adjusts its response accordingly.
In one simulation, the AI‑controlled “red side” coastguard vessel started by assigning a 72 per cent probability that a nearby ship was simply passing through harmlessly.
But when three vessels suddenly formed a wedge and began closing in from different bearings, the system flipped its assessment: the probability of “successive probing” jumped to 68 per cent.


The AI then ordered a water cannon to deter further advances – a measured, non‑lethal response that stayed well within the rules of engagement.
Across dozens of simulated runs, the AI‑powered system never once triggered a weapons‑use violation, effectively reducing the risk to zero. It also cut the overall rate of rule breaches – such as getting too close to another vessel – from 1.2 per cent to 0.8 per cent.
By comparison, the older rule‑based system correctly guessed a vessel’s intentions only 68.5 per cent of the time, while the AI reached 89.6 per cent.
The secret lies in the way the AI thinks. Traditional systems rely on simple if‑then rules: if a boat crosses a certain line, it is flagged as a threat.
The new framework, by contrast, merges radar tracks and other sensor data with a “causal model” – a kind of digital brain that reasons about what a vessel is trying to achieve, rather than just reacting to its movements.
In one test, a group of simulated fishing boats pretended to retreat, hoping to distract the coastguard ship. Moments later, vessels from another direction suddenly accelerated in what looked like a collision attempt. The AI saw through the deception, linking the manoeuvre to a likely trick and holding its ground.
“Our method generates more rational and flexible opponents and effectively exposes plan weaknesses,” wrote Sun Shengzhi, a professor at the China Coast Guard Academy and the study’s lead author.
The academy, which trains officers for the People’s Armed Police Force, published the research on August 10 in the Chinese journal Command Control & Simulation.




The war game was built around a “foreign interference during fishing protection” scenario – a thinly veiled reference to the kinds of grey‑zone confrontations that regularly flare up in the South China Sea and other contested waterways.
One Chinese patrol vessel guarded fishing boats while four to six simulated foreign vessels carried out low‑intensity harassment and reconnaissance. All sides had to obey constraints: a minimum safe distance of 0.5 nautical miles (0.9km), strict weapons‑authorisation protocols and other operational limits.
The central challenge was figuring out what a small, nimble boat really wanted, because in the maritime “grey zone”, hostile probing can look a lot like everyday navigation or fishing.
Sun’s team believes the AI framework can do more than just lower the risk of accidental gunfire. By linking intention recognition, constrained decision‑making and adaptive opponents into a single system, it makes war games themselves more realistic – exposing weaknesses in current plans and sharpening the decision‑support tools available to human commanders.
Future work, the paper said, would extend the approach to more complex, cross‑domain scenarios and explore how humans and AI can make decisions together.
Still, the researchers were careful to note that taking the human out of the loop entirely was not the goal.
“AI decisions remain aligned with higher‑level strategic intentions and ethical norms would require further research,” Sun wrote.


Neurobiologist Ye Bing says his return to China is not spurred by Michigan campus tragedy

China’s stealth fighters, critical minerals from lithium waste: 7 science highlights

China launches its first commercial radio-tracking satellites in electronic warfare leap

China’s cancer breakthrough on CAR-T cell therapy shows progress

Science

Chinese coastguard has 0.3% chance of using weapons in disputed waters, war game suggests

Screenshot from a China Coast Guard social media post showing the Chinese military holding joint naval and air drills in the South China Sea. Photo: The Paper

Chinese coastguard proposes turning South China Sea isles into robotic fortress

Underwater drones, along with their aerial and surface counterparts, could be used to transform defences in the South China Sea. Photo: Reuters

China confirms 2 deaths in Scarborough Shoal crash during Philippine ship chase

One of the China Coast Guard vessels damaged in the crash near Scarborough Shoal last year, according to Philippine authorities. Photo: Handout

Yi Xinyu (left) and Cheng Long (right) have been posthumously recognised by the People’s Armed Police. Photo: Handout

China is sending scientists to Iran for rare earth ‘exploration and processing’

A display of rare earth elements at the National Museum in Beijing. China holds the largest reserves of the metals in the world. Photo: Reuters

China’s glueball ‘triumph’; SpaceX rocket debris caught on video: 7 science highlights

The Beijing Spectrometer III particle detector at the Beijing Electron Positron Collider in Beijing, where researchers have found the strongest evidence yet for the long-sought glueball. Photo: Handout

An image of the moon’s surface shows the estimated impact sites for the SpaceX rocket debris according to Chinese firm ATmoto Technology (pink) and Project Pluto (blue). Photo: QQ.com

Artist’s illustration of Cygnus X-3, a binary system in which a black hole likely feeds on a massive companion while launching near-light-speed jets. Photo: NASA’s Goddard Space Flight Center.

Chinese engineers are developing a two-pronged method to destroy an asteroid on a collision course with Earth using a powerful nuclear bomb. Photo: Getty Images

China is making a great leap forward in food production – this time for real

China’s summer grain harvest this year reached 301 billion jin or about 150 million tonnes, according to the People’s Daily. Photo: Beijing Academy of Agriculture and Forestry Sciences



News

China

Researchers create AI framework to read unknown vessels’ intentions by identifying them as ‘hostile’ or ‘harmless’ to assess risks at sea


Researchers create AI framework to read unknown vessels’ intentions by identifying them as ‘hostile’ or ‘harmless’ to assess risks at sea.


Chinese coastguard has 0.3% chance of using weapons in disputed waters, war game suggests

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Researchers create AI framework to read unknown vessels’ intentions by identifying them as ‘hostile’ or ‘harmless’ to assess risks at sea

Researchers create AI framework to read unknown vessels’ intentions by identifying them as ‘hostile’ or ‘harmless’ to assess risks at sea