Airbus is testing AI in one of the most delicate maneuvers of a flight: landing

Landing a commercial plane seems, seen from the window, to be an almost routine sequence, but in reality, it is one of the most demanding phases of the flight, a maneuver in which pilots, navigation systems, weather conditions and airport infrastructure have to fit together with enormous precision. What Airbus is investigating now is whether the artificial intelligence can help with that fit. Their proposal involves cameras installed on the plane itself and artificial vision to analyze the runway references during landing in real time.

What the company has put on the table is called Vision Landing Application. Airbus has presented it in the context of VivaTech 2026 as a technology still in the research phase, so we are not looking at something that will reach commercial aircraft tomorrow. However, the idea it leaves behind is quite simple to understand: saving all the distances with aviation, it is conceptually reminiscent of what we have already seen in autonomous terrestrial vehicles.

Autoland is not new, but aims to evolve

Here you have to separate two things that may seem the same, but are not. commercial airplanes They can now land automatically under certain conditions, but that does not mean that the system is always available, at any airport and regardless of the crew. They are needed certified aircraft, adequate infrastructure, authorized procedures and pilots trained to operate within that framework. As we can sense, the novelty that Airbus is exploring does not eliminate that reality: it tries to add another form of guidance, born within the plane itself, to an ecosystem where the pilot continues to be a central piece.

Regarding the demonstration at VivaTech, it should be noted that we are not talking about a plane landing in the middle of the event or a commercial test carried out in front of the public. The exhibit was intended to explain how computer vision can improve automated landing procedures. It is less spectacular than imagining an A350 landing at a fair, but much more important to understand where the technology really is.

Now, all this does not come from nothing. Airbus places it within an automation roadmap that began to take shape years ago with ATTOLa project launched in 2018 to explore autonomous taxiing, takeoff and landing using image recognition, without relying on conventional ground systems such as ILS or GBAS. Then other programs arrived: DragonFlyfocused on pilot assistance, automatic emergency operations and workload reduction during taxiing; and Auto’Matewith a different objective, in-flight refueling, but with very close technological bricks, such as cameras, LiDAR, high-precision positioning and AI algorithms.

The next name in that chain is Optimizean Airbus UpNext demonstrator that the company describes as a kind of A350 cabin on wheels. It is not an airplane, but a test vehicle designed to bring sensors, systems and automation to the real environment of an airport without converting each test into a flight. This includes cameras, 4D radar, LiDAR, trajectory protection models, functions against runway incursions and even a virtual assistant to interpret air control authorizations.

Vision Landing Application aims to be useful in at least two especially sensitive cases: remote airfields with little or no advanced infrastructure and environments where GNSS, the satellite navigation that many systems use as a reference, may be degraded, interfered with or directly unavailable. In these cases, the aircraft being able to visually interpret what is in front of it does not replace operational safety, but it does add a potential support network.

The expression Airbus uses is “embedded AI”, but we can translate it more clearly as embedded AI. The difference matters: it is not an AI supported by external servers, but a capability integrated into the aircraft systems. In an airplane there is no excess energy, there is no excess calculation capacity and it is not enough for an algorithm to work well in a demonstration. To get closer to certification, the European manufacturer needs the behavior of the hardware and software to be controllable, traceable and compatible with the safety requirements of commercial aviation.

This is one of the reasons why it is wise to avoid the easy jump to pilotless aircraft. What Airbus describes is much closer to a cabin with better aids than an empty cabin. Its systems seek to alleviate repetitive tasks, improve crew attention and add layers of information. If onboard AI ends up entering commercial aircraft, its first reasonable function will not be to replace the pilot, but rather to give it better tools.

From there to the commercial plane there is still a long way. Airbus will have to demonstrate that this technology works reliably in very different scenarios, integrate it with the rest of the aircraft systems and go through a certification process designed precisely to prevent a promising innovation from reaching real operation prematurely. The Vision Landing Application does not change the way of landing tomorrow, but it does show a very specific direction of where at least part of the industry is heading.

Images | Airbus

In Xataka | Spain had a master plan for the European fighter. The problem is that Germany just got a girlfriend with a lot of “money”