Why would attempting to land using only instruments 'almost certainly' lead to a crash, and why can computers do it safely?

Question

I have no real experience in aviation but I was on the site just out of curiosity. On This question, the top answer states that attempting to land with only the instruments would almost certainly lead to a crash, which surprised me. I understand of course that it would be much more difficult to land without being able to see the airfield, but what factors make such an extreme difference? Also, some computer systems can land a plane, and obviously such computers are using only instruments - what capabilities or information do they have that a pilot doesn't?

As a secondary question, could there be circumstances where it is relatively safe (or at least not catastrophically dangerous) for a human pilot to land using only instruments? For example, would it make a difference if you were landing on a salt flat, completely level and with arbitrary clear space in all directions? Are there certain planes where attempting a manual landing with only instruments might be practical?

Edit: To synthesize/summarize what I understand from the answers so far:

1. Planes capable of automated landing have additional, high-precision instruments that most planes don't have
2. Computers are able to continuously monitor many different instrument inputs and respond appropriately to them all, while a human pilot would struggle to manage so many different streams of information simultaneously

Answer 1

The instruments used typically by the pilot of a light single engine aircraft (such as depicted in the cartoon in that question) are either a traditional sixpack, or at best a glass cockpit approaching the complexity of those seen in airliners.

Traditional

Glass

In both these setups what you have is speed, direction and of course altitude, but that altitude relies on the pilot having correctly set the barometric pressure. While accurate (assuming set exactly correctly) I'm not sure you would trust it to the foot/meter and therefore using it to tell you when to flare for landing would be.... error-prone to say the least.

Airliners, some of which do support full autoland, rely on much more sensitive sensors, with multiple redundancy. More exactly, they will use a radar altimeter to tell the computer exactly how high above the ground it is. This, coupled with accurate glide-slope information provided by Instrument Landing Systems (ILS) or GPS, will allow airliners to much more accurately use their multi-redundant computer systems to effect a landing without pilot input.

It should lastly be noted that the sort of complex equipment that can be added to light, single engine, aircraft means that it is possible to have autoland - at least for emergency situations. Garmin have brought such a system to market in the last few years which uses a traditional form of Intrument Landing System (ILS) not GPS.

Answer 2

There are two key factors to consider: precision, and data processing.

The instruments in the cockpit are only designed to get the pilot "close enough" so that when the runway becomes visible they can land visually. One big challenge is judging the flare so the rate of descent is reduced and the airplane touches down at a reasonable rate. Too early and it could stall and hit the ground too hard, too late and it doesn't slow down enough before hitting the ground. This is harder to do only on instruments, especially in most small planes that don't have a radar altimeter to more precisely measure the height above the ground.

Humans have a hard time focusing on more than one thing at a time and it takes some time to switch their focus. Flying on instruments requires a pilot to be monitoring and controlling attitude, airspeed, altitude, and heading all at the same time. In the air they learn to scan the instruments one at a time and this is good enough for the speed at which things happen there, but at landing things must happen much faster. Even a slow computer by modern standards can switch focus millions of times every second. This lets them constantly monitor all of the sensors and process that information to guide the airplane onto the runway.

To put it another way, humans are most efficient processing visual information and that's why they land looking at the runway out the window. Computers are very efficient at processing numbers like acceleration measurements and radio signals, so they can use that information effectively to control a plane even at landing. Computers are also getting better at processing visual information.

Answer 3

Some planes do have autopilots systems that can land the plane. They know as category 3 auto landing autopilots. to use them the plane has to be certified, testing the autopilot system. the crew has to be certified, training them to use it and know its limitations. Finally, the airport has to be certified. The runway requires precision equipment to send the signals the planes uses to know exactly where the runway is.

Some airports like San Francisco have many times each year where the visibility is 0. Only category 3 auto landing systems can be used. Other airports like Phoenix seldom had low visibility. They would not bother to install the equipment needed.

These systems are expensive and few if any private aircraft will be equipped with them. they are normally only found on large commercial aircraft.

Answer 4

In 1929, Jimmy Doolittle became the first pilot to take off, fly and land an airplane using instruments alone, without a view outside the cockpit. Hence a human can do. There there no computers as such in 1929. About ten years will pass till them appearing even on submarines.

However he used "the most heavily instrumented airplane in the world at the time", including a specialized, very precise altimeter and radio marker beacons.

Hence I think typical instruments as installed in that small aircraft may not be good enough or not optimized for the task, even today.