Chinese military researchers claimed to have conducted a dogfight between a human and AI-operated Unmanned Aerial Vehicle in a real-world condition where the AI defeated the human opponent.
A team of researchers led by professor Huang Juntao of the China Aerodynamics Research and Development Center, a People’s Liberation Army research institute in the southwestern province of Sichuan, published a paper on February 27 in the peer-reviewed Chinese journal Acta Aeronautica et Astronautica Sinica, stating that the AI showed superior performance in cutting-edge close range fighting.
The dog fight entailed two small uncrewed fixed-wing aircraft, one piloted by an AI while the other controlled remotely by a human pilot on the ground.
When the fight began, the human made the first move to gain an edge over the AI-operated drone. However, the AI reportedly predicted his intention and managed to outmaneuver the opponent by making a counter move and sticking close behind the opponent.
The human opponent tried diving into the aircraft to lure the AI to crash to the ground. However, the latter reportedly moved to an ambush position and waited for the opponent to pull up.
After that, the human pilot performed another tactic called the ‘rolling scissors,’ which entails an aircraft suddenly slowing down and changing its course to get the opponent to overshoot in the chase.
Likewise, the fight continued for about 90 seconds, after which it was called off because the human pilot could not evade the AI-controlled aircraft.
Artificial Intelligence To Be The King Of Air Combat?
“The era of air combat in which artificial intelligence will be the king is already on the horizon,” said Huang and his colleagues in the paper, noting that aircraft with autonomous decision-making capabilities “can completely outperform humans regarding reaction speed.”
According to Huang’s team, this was also because the AI did not have to consider human worries while making sharp turns, such as blood being drained from the brain by excessive gravitational pull or concerns about being harmed.
“With superior calculation ability, it can more accurately predict the development of the battle to gain the initiative in the confrontation,” Huang’s team said.
That said, Huang’s team notes that taking AI up in the sky is very difficult compared to running it on the ground because of the limited computing resources on a plane that can severely undermine the AI pilot’s performance.
The researchers explain that the actual conditions in the real world are more complex and unpredictable than those created by mathematical models in a ground simulator. Engineers must also factor in the cost and risks of a crash.
“The problem of aircraft air combat confrontation is very complex with high dynamics, strong real-time characteristics, and a larger solution space,” Huang’s team said. “This poses a huge challenge to the real-life realization of intelligent decision-making.”
The researchers said that there were hardly any reports of bringing the technology from theory to practice, and they built the AI dogfight pilot to simplify it for the Chinese military to use.
For example, the AI’s decision-making process is not related to the hardware of a particular aircraft, meaning the AI pilot can operate on almost any fighter jet in the People’s Liberation Army (PLA) inventory.
US-made Chips Used In Creating The Chinese Artificial Intelligence
Notably, the report about the dogfight between Chinese AI and human-operated drones comes after the US military announced recently that a modified F-16 fighter jet had completed an Artificial Intelligence-controlled test flight for the first time in history.
According to Huang’s team, the research on AI pilot technology began in the US around 60 years ago. However, China has quickly caught up with the US even after starting late.
In 2020, a deep-learning AI system developed by Maryland-based company Heron Systems defeated some of the most experienced F-16 pilots in all five dogfights on a ground-based simulator. And within a year of this, Huang’s team claims to have achieved similar victories in simulations with AI that consumed a fraction of the computing resources used by its US opponent.
Ironically, Huang’s team used an American-built chip to build their AI pilot. The crew installed the Nvidia Jetson TX2 module as the AI brain in the aircraft.
The US government has banned the California-based Nvidia Corp from exporting its most powerful computing chips for artificial intelligence work to China. However, the Jetson-series chips produced in China were not on the list.
The Nvidia chip is said to have allowed the AI pilot to make 1,000 decisions every second based on data collected by sensors on board the aircraft or from the ground.
Are Dogfights Even Relevant Now?
What remains questionable, though, is the relevance of Huang’s team’s AI pilot and their simulated dogfight with the current realities of aerial battles, where experts argue that within visual range (WVR) combat engagements or what is commonly referred to as ‘dogfighting’ has taken a back seat.
Air Marshal Anil Chopra (retired) of the Indian Air Force (IAF), who is currently the Director-General at the Center for Air Power Studies (CAPS), New Delhi, in a recent article for EurAsian Times, examined the statistics of air-to-air combat engagements which showed that dogfighting has become less relevant in the modern era.
Similarly, a study by the United States Air Force (USAF) which analyzed over 1,450 air-to-air engagements since 1965, also found that long-range weapons and sensors have dramatically decreased dogfighting instances.
Huang’s team said in the paper that the proliferation of stealth and electronic warfare technologies had kept dogfighting relevant, which warrants continued research on close-range combat.
“Due to the advancement of stealth and electronic countermeasure technology, 25-40 percent of air combat will be carried out at close range. Research on close-range combat has significant value in practice,” Huang’s team said.
However, in his article, Chopra notes that “both Russia and China are developing low-frequency radars to allow detection of stealth platforms,” meaning long-range air-to-air weapons are not going away anytime soon.
Russian-made radars like the 3D VHF AESA 1L119 NEBO SVU radar could possibly detect very low-observable targets like the US-made F-35 from as far as 152 kilometers.
This is a significant distance, as it could enable a fighter like Su-35 to conduct a cued search and obtain a lock on the American fifth-generation aircraft to fire its long-range air-to-air missiles, such as the Russian Vympel R-37M.