Man vs. Machine: Who’s Boss?

 - January 1, 2013, 2:20 AM

Many pilots had their first–but thankfully second-hand–exposure to the pitfalls of flight systems automation when they watched a remarkable video of an Airbus A320 performing a gear-down, nose-high flypast demonstration at the small French airport at Habsheim in 1988. At the end of the flypast, onlookers expected that the pilot would increase power and the aircraft would climb away. But it didn’t. Instead it continued straight-and-level at between 30 and 100 feet agl (both were claimed in subsequent testimony) before flying into trees just beyond the airport boundary. Three of the passengers died and there were several injuries. The aircraft was mostly destroyed in an intense fire that quickly broke out.

What happened? We may never know for certain, but the consensus among technical specialists and A320 pilots was that the aircraft was most likely under flight system autothrust control–a perfectly acceptable procedure–during the flypast, where power would be increased automatically should the airspeed fall below the speed set by the pilot, which have might have been, say, 120 knots. Manually pushing the thrust levers forward to increase power in autothrust mode would have had no effect. The crew could increase power only by deselecting autothrust.

One reason the cause remains unknown is that, allegedly, the aircraft’s flight data recorder was removed from the aircraft shortly after the crash and was returned to the inquiry many days later. There were accusations that a different recorder was returned. The captain was found guilty of negligence and sentenced to a term of imprisonment. All indications did, however, point to a tragic case of mode confusion.

Almost 35 years later, while mode confusion is far better understood and flight-deck procedures have greatly reduced its occurrence, errors can and still do occur, with some resulting in incidents, although these have remained relatively minor. Nevertheless, the rapid advantages in technology, and the new procedures these will introduce as NextGen, Sesar and other regional plans are introduced, call for increasing diligence in the human/machine arena and recognition that new approaches and new standards are likely to be required throughout the aviation industry.

Systems Manager, not Monitor

For this reason, the agenda at ICAO’s Twelfth Air Navigation Conference (ANC), held in Montreal in November, included the specific area of human performance. Presenters included the ICAO Secretariat; the Civil Air Navigation Services Organization (Canso); the International Federation of Air Line Pilots’ Associations (Ifalpa); and the International Federation of Air Traffic Controllers’ Associations (Ifatca). All agreed that the coming era of new technologies and control systems–in the air and on the ground–will in some cases radically change the ways we measure and incorporate both technical and human performance capabilities.

Until fairly recently, human involvement in an automated environment was viewed as mainly a monitoring function, with little or no authority other than to respond to alerts in a set fashion or procedure. While this remains basically true, Canso proposes that while parts of the system should be automated to help the human, the human must be kept engaged and in control. And, more important, automation should be introduced only to improve overall system and human performance, not just because the technology is available.

Among other recommendations, Canso urges that a balance be achieved between the efficiency created by automation and the human capability to intervene and recover from non-nominal and/or degraded modes of operation–that is, an automation failure strategy. The human must always be in command, Canso emphasizes. The human will be the automation manager, and not the automation monitor, and the automation’s sole purpose will be to assist humans in managing their tasks safely, efficiently and effectively.

Presentations by the other organizations followed the same general lines, with added emphasis on areas peculiar to their roles. Ifalpa, for example, feels that since they are legally responsible for the safety of the aircraft and its passengers, pilots must have the ability to fly the aircraft manually during all phases of flight when the automated systems fail or are unable to maintain a safe level of flight. Note that Ifalpa deliberately stresses when, and not if, such a failure occurs, adding that this certainty of failure bolsters the need for retention of basic flying skills that many feel is threatened by overemphasis on automation.

Ifalpa believes that advances in technology and automated systems must be viewed in the context of the human occupying the central role in the whole system. Pilots must be able to manage every level of automation and be able to intervene manually when required. There must also be a clear indication of the status of any automation function, including the expected future state. Automation is simply a tool, and automation training should enable pilots to correctly identify and predict the actions of the systems and to control them during both normal and abnormal situations. Ifalpa wants to see the planned ICAO Loss of Control Avoidance and Recovery Symposium next year underscore the lack of a fully integrated human-factors approach to current system design.

The fully integrated approach was a theme in all presentations and was aimed at the whole aviation industry, including air and ground system manufacturers. Here, speakers urged system developers to become more cognizant of humans’ need to maintain a continuous understanding of system status, particularly in the light of the much greater complexity of Sesar, NextGen and others where their implementations will include numerous sub-systems or networks of automation supporting a wider community of essential information flows. As Ifatca put it, “The human contribution will be crucial to run the future aviation business in a safe and efficient way, because the human component is the most flexible and adaptive component in the aviation system.”

Training is also critical, with questions about the most appropriate candidates for future employment. For example, are the skills and abilities necessary today adequate to meet the challenges of 2025 or 2030? What qualities do we expect to need in the future? Is there a risk that, say, pilots and controllers recruited today will one day need to transition to a role for which they were not selected? It has been claimed that automation, providing a substitute for experience, will reduce training requirements, but the current attitude is that this expectation is not realistic.

The human performance discussions at the ICAO ANC shed valuable light on a relatively infrequently discussed, yet vital, aspect of our industry. It is to be hoped that this focus will intensify as aviation moves inexorably ever more deeply into automation.

Comments

Iker's picture

This is a real problem because the engineers want to place their creatures (computers) in charge as a matter to show their power in modern society.
Unfortunately this will create more unemployment (read more here : http://tiny.cc/4mq8pw) while in reality there is nothing worse than a computer created by a human being. Since, they are made by humans, computers have defects and problems that can sometimes be really serious.

check the math's picture

Editor mistake on down?

Show comments (2)