Plane logical

With human error blamed for the majority of air accidents is there now a case for fully-automated aircraft?

Last weekend’s crash of a China Airlines plane in Taiwan, which killed all 225 people on board, has heightened public anxiety over the safety of air travel. The 22-year-old aircraft disintegrated in mid-air and crashed into the Taiwan Strait, 20 minutes after take-off from Taipei, leading to fresh criticism of the airline’s poor safety record.

Public confidence in flying took a huge knock last year after September 11. But the tragedy was also a catalyst for some long-overdue questions to be asked about air safety. The slump in passenger numbers following the attacks halted, temporarily, a trend that was beginning to alarm industry experts, particularly the US Federal Aviation Administration.

If air travel expands in line with predictions and the rate of crashes remains at the level it is today, major accidents are expected to rise to almost one a week within the next 20 years. So the question remains, how can planes be made safer?

Last week it was claimed that UK air traffic managers are having such problems reading the screens of their new control system they are sending aircraft into sectors at the wrong altitude – and even having difficulty distinguishing between the codes for Glasgow and Cardiff.

This means that aircraft with the ability to plan ahead and avoid other planes could prevent tragedies caused both in the air and from the ground. Although it looks unlikely that the China Airlines crash in Taiwan was due to pilot error, a large number of accidents are. Perhaps most tragic of these was at Kegworth, Leicestershire in 1989, when a Boeing 737 plummeted on to the embankment of the M1, killing 47. The aircraft crashed after the pilot reacted to a fire in the left engine by mistakenly shutting down the right one, despite a large amount of information on the plane as to which was at fault.

Better training and information for pilots may help, but not enough to reduce accidents by such a large amount. So the only way forward for the industry is to use technology. According to David Allerton, professor of avionics at Cranfield University, we are rapidly reaching the point where computers and sensors will be far better equipped than pilots to fly aircraft. ‘The difficulty is that pilots, particularly under stress or heavy workloads, make mistakes,’ he says.

Allerton believes the next generation of aircraft will be completely autonomous. Although systems are not yet in place to cope with this change, it will simply mean an extension of existing technology, as aircraft already carry an enormous amount of on-board processing power, as well as sensors to take measurements such as speed and altitude. ‘You need more sensors, more processing power, and more information (if you continue to have a flight crew at all) to reassure them they are where they think they are, and are not about to fly into a mountain or bad weather.’

Progress towards this may come in small steps, such as using autonomous aircraft to carry only freight for five or 10 years, until they are proved to be safer and the public is confident they are safer than piloted planes. Another option would be for passenger aircraft to fly totally autonomously, but with a member of the flight crew on board, sitting in an ordinary seat with a computer and able to override the system and fly the plane if necessary.

Some question whether computers can be relied upon to fly safely without picking up potentially catastrophic glitches, particularly given the recent problems with the new computer system at National Air Traffic Services. But all on-board aviation systems are designed with the possibility of failures in mind, with each aircraft fitted with back-up equipment as a fail-safe measure, says Allerton.

‘We have this Apollo 13 mentality, where because there are two pilots on board, we think they will be able to get out their screwdrivers and change the fuse and we will all be saved. But in practice they will take out the wrong fuse and we will all be killed – that is the evidence,’ says Allerton.

In any case, pilots already do very little in terms of manually flying the aircraft on a normal journey. Once the pilot has taken off the autopilot takes over and flies the aircraft to its destination. If the destination airport is fog-bound the autopilot will perform the descent, approach and landing, and bring the aircraft to a halt at the end, says Allerton. From here the pilot takes control again and taxies the aircraft back to its runway position.

‘Taxiing along a straight line is not rocket science for computers. Nor is making them go in a straight line and take off, rotate and rise to 200ft and raise the undercarriage, or taxi back to the runway position at the other end. We already have that technology largely in place,’ he says.

Work is already under way to make aircraft computer systems more autonomous. This can be seen particularly in air traffic management, which is soon to move to data links rather than the less reliable and slower voice communication. This will eventually mean on-board computers do much more than simply follow a flight path, says Allerton. ‘The on-board computer will be able to say what the aircraft is, where it is, the heading, speed, and its intentions for the next 15 minutes.’

The computer will be able to communicate this information to other aircraft and the ground, allowing planes to take early action to avoid each other where necessary, rather than relying on their on-board Air Traffic Collision Avoidance System, which only has a range of a mile.

‘This will mean much better separation assurance and safety, and improved performance: aircraft will be able to fly in straight lines from A to B, they will not have to zigzag to avoid certain points.’ It should also mean they no longer have to circle the runway when they reach Heathrow, as they will already have sped up or slowed down by the small amount necessary to put them exactly one mile behind the plane in front. In this way traffic flow can be increased to meet rising demand for air travel, without risking safety.

The move towards more autonomous aircraft can be seen in research being undertaken to develop avionics systems that allow them to fly at night, and in all weather conditions. AgustaWestland, the joint venture between GKN and Italy’s Finmeccanica, is working on a system that enables helicopter pilots to fly in full white-out conditions. Information gained from various sensors on the aircraft, including radar data, height readings and air speed, will allow it to fly around obstacles such as power cables and other aircraft, says Dean Goff, senior engineer for simulation systems at the company.

The information gained from the sensors is collected by the computer system and combined with data from the on-board GPS, Traffic Collision Avoidance System and terrain databases to determine the location of any obstacles.

But unlike simple obstacle-avoidance algorithms, which the company’s engineers have been working on in the past, the Navigation and Obstacle Avoidance for Helicopters (Noah) system does not simply find the straightest path around these obstacles, says Goff. ‘The straightest path is not necessarily where the pilot wants to go, as there are also issues such as the rules of the air to consider.’

The system searches along a series of possible paths from its position, and weighs each route against issues such as obstacle avoidance and the rules of the air, to find the best path.

As parameters can be entered into the system before each flight, including speed and altitude, Noah can also take into account the pilot’s individual preferences. ‘There are several different ways of flying, and the priority on a flight might be to use the minimum amount of fuel, while if it is cloudy the pilot might want to fly at a lower altitude,’ says Goff.

Once the system has chosen the best path around obstacles such as military airspace zones, it displays this route to the pilot. ‘It would be particularly useful in search-and-rescue operations in mountainous terrain, where the pilot has to fly as low as possible to search, but does not want to hit obstacles such as pylons.’

But Noah can also go a step further. AgustaWestland is using its experience in making helicopters to feed information on the capabilities of each aircraft into the system, says Fergus Crawford, principal engineer. ‘We are moving towards more sophisticated flight path guidance, which is more intelligent and flies more realistically,’ he says.

As well as knowing what obstacles must be avoided, the system would also be aware of the particular helicopter’s capabilities in certain weather conditions. ‘Noah might say: OK, we need to fly to this destination by this route, and tell the pilot he or she has to change velocity, height or attitude in order to get there.’

Having developed a system that is able to plan its own routes despite bad weather conditions, rather than simply following a flight plan, it would be just a short step to using autopilot to fly the aircraft itself, says Prof Chris Harris, who was involved in the early stages of the research project at Southampton University. ‘My personal view is that pilots should not be flying the aircraft, it should be an autopilot system,’ he says.

As well as reducing short-term demand for air travel, September 11 sparked a great deal of interest in avionics systems capable of overriding a pilot’s commands and flying around buildings. This can be done, says Don Bateman, chief engineer for flight safety systems at Honeywell. But technology is more likely to be of benefit in preventing pilots from making mistakes and flying into terrain such as mountains, he says. ‘If you want to make an aircraft resistant to flying into a mountain, water or a building, you simply make limits on the system’s controls to resist that. So a pilot could fly anywhere else they want to go, but not into that area.’

Honeywell is developing displays capable of detecting and highlighting the aircraft’s altitude in relation to local terrain. With its Enhanced Ground Proximity Warning System, if the aircraft is flying at a normal height above the ground, the display will not show anything, but if it dips slightly low, the display will change to a dot matrix of green, and then yellow. ‘If the aircraft continues to decrease in altitude, the display will change to an intensive dot matrix of red, and then become a solid yellow as a warning. That is very powerful: if a pilot saw it, they would know instantly that something was wrong,’ says Bateman.

As well as warning the pilot, the technology could take action to ensure the aircraft is not flown into the ground, he says. ‘We do not want to surprise the pilot, but we have had accidents where despite advance ground proximity warnings the pilot has done the wrong thing, when they have become confused by what they have seen. So if there was a warning and the pilot did not respond, the aircraft would resist.’

The company is also working on displays to improve the crew’s awareness of precisely where the aircraft is, if they cannot see easily due to darkness or bad weather.

But however much technology is installed on an aircraft, where there is a human element involved there is always the risk of error. Pilots make mistakes because they get tired and bored, or stressed and confused.

And with air traffic set to increase again over the next few years following its recent slump, a rise in the number of pilot errors, with potentially fatal consequences, will inevitably follow. The idea of an autonomous aircraft might seem outlandish now, but in years to come we might be equally nervous about boarding a flight that is piloted manually.