不介意我多嘴确认一下可以用中文评论吗？我记得管理员好像是说过这里只能用英文，毕竟我几乎没有看到有写中文的。能用双语最好：）
Self Propelled Gun/Rocket Launcher
- Thread starter FORBIN
- Start date
好像是只能用英文写评论，但可以用中文回答一些简单的问题。
Please use English on the forum.
Yes, please provide an English translation if one is writing in Chinese, as this is an English language forum.
Obviously we don't have anything against writing in Chinese, but for regular forum talk (as opposed to conveying Chinese language information here that then gets translated), it is better to write primarily in English.
China tests AI-powered long-range artillery that can hit a person 16km away
Multiple test launches achieve lethal accuracy greater than guns used now
High-precision cannons could significantly cut costs of war, researchers say
Stephen Chen
Published: 7:00pm, 17 Apr, 2023 Updated: 7:00pm, 17 Apr, 2023
China’s military has used artificial intelligence to improve the accuracy of long-range artillery and potentially reduce the cost of warfare, according to a team of scientists working on the technology for the People’s Liberation Army (PLA).
In multiple tests conducted under various conditions last July, the researchers determined that their AI-powered laser-guided artillery could hit human-sized targets 16km (9.9 miles) away.
The precision achieved in the tests, which exceeded expectations, was far higher than that of any big guns in service, according to photos of the tests that showed the shells hit target boards in the bull’s-eye.
“Artificial intelligence is evolving quickly. More researchers are applying the technology to trajectory planning problems,” said the project’s team leader, Professor Wang Jiang, from the Beijing Institute of Technology, in a paper published in the peer-reviewed journal Acta Armamentarii on April 6.
Traditional artillery shells can land 100 metres (328 feet) or more away from a target. Guided artillery shells, which can make course adjustments during flight, are finding wider applications by militaries in China, the United States and other countries.
However, their accuracy has been limited due to the huge amount of real-time data that must be calculated using traditional mathematical models. Flight variables such as wind, temperature and air pressure can limit the precision of an artillery round to the point that it could miss its target by a few or tens of metres.
Instead of traditional mathematical methods, AI offers the potential for faster data-processing speeds, according to Wang, as well as collaborators from China’s defence industry and experts from a Beijing-based China-United Arab Emirates Belt and Road joint laboratory on intelligent unmanned systems.
Once a smart shell is launched, it must quickly collect and analyse a wide variety of environmental data to fine-tune its course – calculations that can increase exponentially with the number of variables.
The shell’s computer chip must be as simple as possible as it must withstand the enormous heat and shock of artillery fire. Facing such demands, processors must often discard valuable raw data to complete the calculations in time, thus affecting overall accuracy.
But with AI, even a slow computer chip can finish necessary calculations using nearly all available data, according to Wang’s team.
As it learns from training based on data collected in real flights or laboratory experiments, AI can bypass some of the more demanding calculations performed under traditional approaches, according to the researchers.
The team also tested several AI models on tasks associated with sophisticated trajectory adjustments during flight. This division of labour among the AI models enabled further accuracy improvements, they said.
Both China and the US are racing to develop smart artillery to help cut the costs of warfare. Artillery shells are usually much cheaper than missiles, and can be quickly produced in large numbers.
Last year, the US Army awarded a US$66 million contract to arms maker Raytheon for an unspecified quantity of GPS-guided artillery smart munitions, with ranges of up to 40km, according to some media reports.
Chinese state media last year released footage of a live-fire exercise, which showed a moving car being destroyed by a smart artillery round, but the effective distance and accuracy of the weapon was not disclosed.
A new smart mortar deployed by the PLA has also reportedly achieved hits within centimetre-precision. However, mortars usually have shorter ranges and lower speeds compared to artillery.
Meanwhile, debate has continued over whether artillery fire requires such high accuracy.
Since shrapnel from a heavy shell blast can hit a person several hundred metres away, some military experts have said precision is unnecessary.
But others have argued that a high-precision cannon would be very useful for the PLA in the event of an attack on Taiwan.
In urban warfare, AI-powered artillery could neutralise enemy units or vehicles hidden in buildings with more efficiency than traditional fire power and at lower costs than missiles, said a Beijing-based defence industry engineer who was not involved in Wang’s project and asked not to be named due to the sensitivity of the topic.
“It will help reduce civilian casualties and damage to surrounding buildings. It will make reunification and reconstruction after the war easier,” he said.
I think that article is not-accurate for several reasons.
1- Guided arty shell is an existing tech. You can achieve pinpoint accuracy using laser guidance or GNSS.
2- "Once a smart shell is launched, it must quickly collect and analyze a wide variety of environmental data to fine-tune its course"
Well, they don't. Both laser and GNSS guidance methods require very little data collection by the munition.
3- "However, their accuracy has been limited due to the huge amount of real-time data that must be calculated using traditional mathematical models. Flight variables such as wind, temperature and air pressure can limit the precision of an artillery round to the point that it could miss its target by a few or tens of metres."
This has nothing to do with smart shells. Guided munitions correct for deviations continuously. There are multiple ways of determining the correct trajectory and staying on that. At no point, smart or traditional shells collect this data. Firing artillery contingent does. The article is talking about traditional shells on this part. Then it suddenly jumps to guided munitions and applies the same logic there. Then it claims artillery shells can only have limited computing power, so these calculations should ideally be removed from the shell. This is simply not true. Existing guided arty shells don't care about atmospheric variables at all after getting fired and they are very accurate.
4- "Both China and the US are racing to develop smart artillery to help cut the costs of warfare. Artillery shells are usually much cheaper than missiles, and can be quickly produced in large numbers."
Not really. Guided arty shells are expensive. And they have little benefits in most situations because gun artillery is quite accurate even without guidance. And there is no race. Both sides have those.
AI could be used for making traditional shells more accurate by improving gun laying in regard to atmospheric variables. And that’s what I think the article tried to talk about.
From my reading, I had a bit of the opposite understanding. I thought your point 2 was an example of the hypothetical AI-smart shell (really, just putting in some real time processing)
It is a lot of buzzword stuff, but last Zhuhai airshow there was an Imaging IR shell (GP155E), so maybe it is kind of related
2022 Zhuhai Airshow Land Systems
www.sinodefenceforum.com
An imaging IR would result in truly exceptional accuracy and would benefit from AI. I wonder if the PLA has any.