Skip to main content
Hamburger Menu Close



Exercise Forging Sabre: Apache, fighter pilots get enemy data faster with help of AI

Exercise Forging Sabre: Apache, fighter pilots get enemy data faster with help of AI

An Apache AH-64D before launching for a night mission. (Photo: Aqil Haziq Mahmud)

BOISE, Idaho: Soaring silently in the sky, the Heron 1 unmanned aerial vehicle (UAV) spots three moving vehicles below suspected to be enemy targets. 

The UAV feeds real-time video back to a big screen in the command post. Commanders there immediately see red rectangles appear around the vehicles. This is the Automatic Target Detection (ATD) system confirming they are threats.

Three F-16 fighter jets are scrambled. They screech towards the enemies, on a mission to take them out. The pilots wait for instructions.

Back at the command post, a different image appears on the big screen. This time it's the Target Look-Ahead (TLA) system showing three symbols moving on a road, their direction clearly indicated.

The system automatically picks an ideal spot for an airstrike, close enough to the jet fighters to be strategic, yet far enough from structures to reduce collateral damage. The spot is denoted with a green box.

The system starts a countdown next to the green box on the screen, indicating how much time the enemy vehicles need to get there. Commanders use this information to precisely tell the F-16s when to move in. Right on cue, the fighter jets intercept and destroy the vehicles, exactly where they’re meant to be.

An F-16 taking off. (Photo: Aqil Haziq Mahmud)

The Singapore Armed Forces (SAF) are using the ATD and TLA systems, integrated within the command post, at this year’s Exercise Forging Sabre in Idaho. The ATD is being trialled for the first time at the exercise; the TLA used this year is an improved version building on previous editions of the exercise.

On Monday (Oct 7), senior exercise commanders described how the artificial intelligence (AI)-aided systems might be used in a real-life mission, stating that the systems have made the decision-making process faster, easier and more precise. 

Exercise air director Colonel Aldrin Tan at the command post. (Photo: Aqil Haziq Mahmud)

“These processes actually help all the battle staff come to a faster appreciation of the situation and recommend those decisions up to me,” said Colonel (COL) Aldrin Tan, 41, the exercise air director who also heads the command post. “So, it enables the entire command post to think and decide faster.”

Defence Science and Technology Agency engineer (information) Joshua Lim at the command post. (Photo: Aqil Haziq Mahmud)

With speed the key in winning such battles, the Defence Science and Technology Agency (DSTA) – which helped develop the ATD and TLA – ensured the systems could complete the complex, multi-step process in a matter of seconds.

“We are able to run it at a real-time rate,” DSTA engineer (information) Joshua Lim, 24, said. “We cannot have any delay in the video because the commanders have to make a very timely decision.”


But it wasn’t always this quick and easy.

Before ATD, command post personnel had to manually analyse congested UAV video feeds to confirm if suspected targets were indeed enemies. For instance, they would have to know what a typical enemy truck looks like, then pick these out in images that might also contain civilian trucks.

READ: Exercise Forging Sabre: Fighter jets to fly longer missions as new tanker aircraft makes debut

The ATD overcomes this by using machine learning. Developers feed the system with intelligence on how an enemy looks like, and the system learns how to recognise it quickly, much like facial recognition. 

An RSAF Heron 1 unmanned aerial vehicle parked at Mountain Home Air Force Base in 2019. (Photo: Aqil Haziq Mahmud)

Still, the ATD is limited by the size and appearance of a target object on the UAV video feed. So if an enemy truck is too far away or well-camouflaged, the ATD might not be able to pick it up.

“But that also means that the enemy is now forced to do so many things on every single vehicle that he has, which increases demand on him to avoid detection,” COL Tan said.


As for the TLA, command post personnel who didn’t have it had to guess the time enemies would need to reach a strike location, based on the distance the enemies had to cover and the speed they were travelling at.

“It’s so prone to errors that it was something that they struggled with,” COL Tan said.

The Heron 1 unmanned aerial vehicle taking off. (Photo: Aqil Haziq Mahmud)

In contrast, the TLA uses AI to calculate the time needed, and even recommends a strike location based on factors like road networks and the blast radius of a chosen weapon, not unlike how an online map suggests a recommended route.

“The computer does it for them,” COL Tan added, giving an example of how an ensuing conservation between the command post and fighter pilot becomes more precise.

“Okay, it’s 45 seconds (until the enemy gets there). You are clear to run in now, and clear to point your weapon in 20 seconds. Because you know that weapon will take 30 seconds to impact.”


Pilots of the F-16 and Apache attack helicopter, two strike assets participating in this year’s exercise, recognise the importance of getting quicker and more precise information from the command post, especially as they use this data to fight together with sensors like the Heron 1.

F-16 pilot Captain Nigel Wong (left) speaking with an Air Force engineer. (Photo: Aqil Haziq Mahmud)

“Obviously, integrating with the other forces has tactical benefits,” F-16 pilot Captain (CPT) Nigel Wong, 27, said, adding that operators of different platforms must first recognise each other’s capabilities and limitations.

Air Force personnel loading weapons on the F-16. (Photo: Aqil Haziq Mahmud)

For instance, the Apache’s less explosive weapons make it a more ideal strike asset than the F-16 in confined urban spaces, where collateral damage should be kept to a minimum. The Heron 1’s targeting laser might also work better than the F-16’s, which can be hindered by weather and terrain.

There are five Apaches participating in this year's exercise. (Photo: Aqil Haziq Mahmud)

“Once you’ve done that, we hash everything out in the planning process, and then finally we (ensure) we are well-deconflicted in that area so that we can effectively and safely get our bombs on target in the whole sense-to-strike process,” CPT Wong added.

Apache pilot Lieutenant Ong Yan Zhuan (right) shares a lighthearted moment with Air Force engineer A Surya, who maintains the helicopters. (Photo: Aqil Haziq Mahmud)

Apache pilot Lieutenant Ong Yan Zhuan, 26, said: “From an operator’s standpoint, getting more precise information is definitely going to aid me in terms of mission panning and such. So with that, we’re able to carry out our mission more effectively and efficiently.”

Personnel at the command post work through a live mission. (Photo: Aqil Haziq Mahmud)

Nevertheless, the exercise air director COL Tan pointed out that the AI systems are meant to help, not replace, command post personnel. Commanders still make the final decision on whether a target is an enemy, and following that, the ideal location for an airstrike.

“I think the business of warfare still requires that humans make that final determination,” he said.

Source: CNA/hz