China’s AI used an clever tactic to win a simulated Mach 11 dogfight

Researchers in China say their Artificial Intelligence (AI) is the first to simulate an aerial battle involving a hypersonic aircraft flying at Mach 11, or 11 times the speed of sound, and the technology came up with a surprising tactic for defeating an enemy. In the computer simulation, a hypersonic aircraft encountered an enemy fighter jet flying at Mach 1.3, close to the top speed of an F-35.

The hypersonic plane’s pilot was then given an order to kill. Instinct, along with the results of earlier simulations, should have directed the pilot to head straight towards the target.

But the pilot, guided by AI the team developed, flew to an unexpected position far ahead of the enemy aircraft and fired a missile “over the shoulder,” tossing it backwards towards the foe.

The missile hit the enemy fighter, which was 30km (18.6 miles) behind the hypersonic plane, at a speed of up to Mach 11, ending the battle in less than eight seconds, according to the computer simulation.

This counter-intuitive approach offered the longest possible killing range with the lowest risk to the pilot, according to the team led by associate professor Liu Yanbin with the Nanjing University of Aeronautics and Astronautics.

The researchers simulated the battle at speeds ranging from Mach 5 to Mach 11. They published the results in the Journal of Beijing University of Aeronautics and Astronautics last month.

Reusable hypersonic aircraft have many potential advantages in air combat, including relatively low mission costs and the ability to fly at extremely fast speeds with high manoeuvrability. However, fire-control systems for launching weapons have not yet been developed for aircraft flying at hypersonic speeds of Mach 5 and above.

Hypersonic flight requires quick response times and puts new demands on fire-control systems, which must be able to perform extremely precise calculations. Liu’s team said their AI could be used in hypersonic aircraft to improve combat performance.

“Artificial intelligence has very broad application potential in this field,” the team wrote. According to the researchers, the most unexpected outcome of the simulation was that at Mach 11, the maximum attack range occurred with the enemy right behind the hypersonic plane’s tail.

In a typical aerial dogfight at speeds below Mach 5, opponents generally try to take each other head on and avoid being trailed by a rival.

“This shows that when the [hypersonic aircraft] is performing air-to-air missions, it can attack from outside the combat theatre by launching weapons at a long distance in front of the target using over-the-shoulder launch,” the team wrote.

After firing the missile, the hypersonic aircraft would quickly leave the battle. This method, with its ability to strike quickly from a long distance, would take full advantage of the hypersonic aircraft’s flight performance and greatly improve the crew’s chances of survival, the researchers said.

“The future air battlefield situation is becoming more and more complex, with combat tasks becoming more difficult. Pilots need to quickly process a large amount of information and make optimal tactical decisions,” Liu’s team said.

“The application of artificial intelligence technology in the airborne fire-control system can improve the aircraft’s situational awareness and the system’s offensive and defensive response speed.”

The team said that in the next step of their research, they would use AI to help multiple hypersonic aircraft coordinate and perform a “multi-wave, multi-task attack”.

“Future strategic deterrence depends on global rapid-strike capability and the depth of penetration. In recent years, all military powers have strengthened their research on high-speed combat aircraft,” they wrote.