Former Shadow 200 TUAV from the US Army pilot here. Not only are fully automated landings feasible, they are happening in practice. The Shadow has a piece of equipment called the T.A.L.S. (Tactical Automated Landing System). Landing a shadow (I did this in Iraq as well mind you) involved aligning the AV in a small box on the screen which turned it green. Provided the airspeed was right and crosswinds were in tolerance, you click the "Land AV" button and the landing is 100% hands off. If at any point during the tals auto-landing something looks off, you push the abort button which powers the engines to full, gains altitude, and lets you circle around to try again. This tech is around and battle tested.

My thought on why the predator / reaper aren't is simply so an officer can go through flight school and get his cool flight wings in the USAF. There is no good technical reason as far as I know landings can't be fully automated with larger UAVs.

I helped develop that technology, glad to hear of someone on HN using it. =))

Did you all read some of the problems which arose? In one crash report, the drone took "off without permission from the control tower". In another, "an armed Predator suffered an electrical malfunction that sent it into a death spiral." A third was partially due to a "melted throttle part." A forth started when "[t]wo minutes after takeoff, the engine failed."

Another concern is that air control gives "priority to passenger planes and order[s] drone pilots to keep their aircraft circling overhead even when they are dangerously low on fuel."

Yes, some of the problems were human error. But from the statistics given, UI errors and improved navigation support give at most a 30% reduction in the crash rates. While what's needed is several orders of magnitude improvement to make it match the commercial crash rates.

I did read the article. I was responding to a post that highlighted the human factors in the conflict.

If the UI was simple, the landing scheduling would be handled automatically. If the plane had too little fuel, the planning failed. Keeping a schedule isn't difficult, human interaction and ill-defined schedules makes things difficult.

I am aware that a central african state isn't exactly the cornerstone of air tech, but good schedules can still be negotiated and kept.

It's like the air france crash where the pilot stalled the plane because of lack of feedback. UI/skill problem.

My argument is that improved UI doesn't appear to be the source for major improvements in reducing the number of drone crashes. While a simple UI might handle some engine failures, it can't handle all hardware failures.

The landing scheduling depends on the local air traffic controllers, which aren't under US direction and can't be solved with an improved UI. The planning problem might be that the local ATC doesn't care for the US military traffic in the area, and this is a form of passive protest. If so - and that is a pure hypothetical as the article doesn't go into those details - then the planning problem is at the high-level political level, and perhaps extending to how the world perception of how the US is dealing with the GWOT. That's extremely far from UI issues.

As minor points, neither Djibouti nor the Seychelles are a central African state. Also, the initial problem with AF 447 was inconsistent airspeed inputs from the pitots, causing the autopilot to disengage. I don't think it's useful to attribute the problem only to "lack of feedback." Yes, there's a "skill problem", but if you continue the analysis chain then there's a training problem, and if you go further then there was an incomplete understanding of the risk model.

It is insane to build a craft with powerful weapons and not reserve guaranteed capacity for landing spots, including backups at other times and locations from main flight plan.

Modern commercial airliners do not land themselves without human input. Someone needs to dial in a whole bunch of parameters then monitor the instruments carefully.

Autopilots aren't infallible; they're basically glorified cruise control with three axes of freedom rather than one.

(I will grant you that commercial aircraft with ILS can land in zero visibility conditions, but that's a very long way from being able to fly themselves to an arbitrary destination without human direction.)

Understandably, commercial airliners do not land via computer because human pilots are better than the autopilots in most commercial aircraft. To say that this will always be the case, especially in an entrepreneurial software forum, is ridiculous. Autopilots will get better and better, and drone pilots are not currently as good as commercial aircraft pilots.

Because... you just argued that people are required to enter parameters, and monitor sensors, and, well, people aren't very good at those things.

[http://www.x-plane.com/x-world/hardware/seeker-avionics/]

i think your parent comment miswrote a period for a comma, so you aren't disagreeing.

Automatic landing of drones is and has been a reality for a long time.

http://www.airliners.net/aviation-forums/tech_ops/read.main/...