Running head: Autonomous Commercial Aircraft
Autonomous Commercial Aircraft: Development and Implementation
Ron Marish
Embry Riddle Aeronautical University
2012
Abstract
The subject of this paper is the development and implementation of autonomous commercial aircraft. These aircraft will routinely fly scheduled routes carrying passengers and cargo with no human pilot onboard. This is the natural progression of commercial flight as a result of the rapid growth of the flying population, the increasing complexity of the national airspace system, and the need for a higher degree of precision and reliability than a human pilot can offer. Automation and artificial intelligence provide these requirements, and have capabilities which can continue to advance exponentially faster than human evolution. This allows these technologies to outperform their human counterparts in information processing as well as their ability to perform complex tasks quickly and accurately.
Autonomous Commercial Aircraft: Development and Implementation
When asked if they would feel safer with or without a pilot on board, most airline passengers would probably respond: “with a pilot”. This would seem to be a natural response, given that the pilot is thought of as the ‘sole manipulator of the aircraft controls’, and the one person that prevents the passengers from certain fiery death - a powerful and justifiable argument. Upon learning that the majority of flying is actually performed by the autopilot and onboard computer systems, most people who stand behind their initial response. However, if the flying public was first educated on the statistics involving fatal crashes and human error, they may have a different answer.
Computer systems and electronic components perform to certain specifications and design limitations that can be cataloged and predicted with a high degree of accuracy. As technological advances continue and discoveries are made, this level of accuracy continues to increase. Conversely, human capabilities are very difficult to predict or catalog with any accuracy, and countless factors affect these capabilities such as emotion, motivation, fear, and panic. In addition, evolution allows humans do advance in capabilities as a fixed rate, which is exceedingly slow in comparison.
Humans have already reached and surpassed the limits of what can be accomplished without the aid of technology. Evidence of this can be obtained from examining the industrial revolution, and how life began to change immediately after. Building better machines and devices has always and always will improve the lives of those who develop them. Taken to the extreme, theoretical physicist Michio Kaku predicts that one day in the distant future, we will all become ‘pets’ of our inventions, which will perform all mundane tasks while we lounge around and enjoy the finer points of life.1 Stepping aboard a fully automated transportation system, that will effortlessly whisk us to the far corners of the globe seems like a good example. As long as we continue to innovate, there is no limit on what can be accomplished.
Accident data shows that human error, and specifically pilot error, is usually a causal factor in fatal aircraft crashes. This chart shows the percentage of accidents from each category of failure.2 Pilot error is clearly the dominant factor in plane crashes over the last 60 years.
Breaking this data down further, some primary causes of pilot error which have led to crashes in the past can be highlighted. They include:
• Flying in poor weather or knowingly flying into a storm
• Controlled flight into terrain resulting from a navigational error
• A pilots failure to correctly interpret cockpit displays and instrumentation
• Failure to execute proper control inputs during a stall
• Pilots failure to inspect fuel levels and running out of fuel during flight
• Improper use of flaps
• Loss of spatial awareness
• Pilot failure to maintain proper speed
• Flying at the wrong altitude, or on altitude other than assigned by ATC
These are just some examples, all of which could be eliminated by automation and sufficiently advanced computers. Of course the argument can be made that computer systems also fail. This is true, however they fail at a predictable rate, and their reliability can be improved indefinitely.
Much of the required groundwork is already in place. Complex autopilot’s and flight management systems, satellite navigation, and precision approaches that can be flown entirely automatically. With some minor improvements, these systems will be the foundation for pilotless aircraft. In some accidents, tragedy could have been avoided if the pilots would have only allowed the computer systems to remain in control. Pilots are subject to disorientation, leading them to believe the instruments are faulty when they are not. Computer systems are not subject to these perceptual errors. Computer systems also have much better reaction times than their human counterparts and in a situation where milliseconds can make the difference between life and death, automatic systems may be the wise choice.
Autonomous drone aircraft are already being flown all over the world. They can take-off, conduct their mission, and land, all with no intervention from a human other than installing the initial program. The technology to safely conduct these flights already exists, though it must be refined before human passengers will be routinely flown. AI systems that can think and make decisions on their own during flight need to be improved upon, because of the unpredictable nature of flying long distances and frequency with which drone aircraft loose communication with the ground stations operating them. The key point to remember is that we already have the knowledge of how to make this work.
For example, landing an F-18 on the deck of a moving aircraft carrier is regarded by many pilots as the most challenging and difficult task in aviation today. According to Mary Cummings, an automation expert at MIT, the automated landing systems on the F-18 always do a better job than the pilots on carrier landings.3 It was simply better at making the thousands of minor adjustments necessary to put the plane in the exact location it needed to be in. These technologies will continue to trickle down to the civilian world, and will eventually dominate all of aviation. Advances in artificial intelligence will hold the key for these improvements, as well as advancements in high speed data transmission.
It is true that drone aircraft are not perfect. The very name ‘Drone’ implies the lack of intelligence many of these aircraft have. The more advanced autonomous aircraft are more correctly referred to as UAV’s, or unmanned aerial vehicles. Drones and UAV’s are known for numerous crashes in recent years; however this is not surprising when considering their explosive growth. The FAA is struggling to keep up with the pace of this growth, and currently technological advances are drastically outpacing the regulatory agencies ability to create new rules governing their flight.4
UAV crashes are not immune to human error at this early stage in their development, and many crashes can be attributed in some way to human factors. The majority of UAV accidents however remain to be directly related to hardware failures.5 An important correlation can be drawn here, as this is a reversal of the current problems with commercial aircraft. In the early days of aviation mechanical failures caused the majority of crashes; that is until that hardware was improved to where it is today. Now, the aircraft are so reliable that the pilot is the primary factor of accidents. UAV’s are only in their primal stages of development, and when the hardware and mechanical systems they use improves only a little further, there will be no weak link in their operational chain.
Hopefully by now the need for autonomous passenger aircraft is apparent. The technology exists, and with some refinement can be safe enough for widespread implementation. The first step will be getting people to go along for the ride. This may not be difficult in countries like China, where massive growth in aviation has caused the national airlines to have a notoriously poor safety record because of lack of well-trained pilots and substandard maintenance.6 In other countries, where current levels of safety are already quite high, it may be somewhat more difficult.
The national airspace system is becoming a very congested place, and future growth is estimated to strain the aging regulations and methodologies currently in use. This is one goal of NextGen, the FAA’s answer to meeting future traffic demands without interrupting service and causing frequent delays. NextGen will reduce spacing requirements between aircraft, and it is able to do this because of automation and better surveillance. The trend of taking more and more duties away from the pilot continues, and NextGen has many components that will assist pilotless aircraft. It will also aid in keeping them safely away from piloted aircraft during the long transition to fully autonomous flight.
Once the flying public accepts the idea of riding on these aircraft, the next step will be demonstrating that it can be accomplished safely. First we will need to develop the cockpit hardware which will interface into the auto pilot systems already in place. As many aircraft already have the ability to land themselves on precision approaches, easily the most complex aspect of any flight, only slight advancements need to be made. More fail-safe and backups of existing systems should be sufficient. Adapting automation to the take-off and cruise phases of flight should be relatively easy, and a reliable GPS signal and attitude reference is already widely available. These systems will guide the aircraft along its predetermined flight path, while Artificial Intelligence algorithms sort out any problems encountered along the way.
Dealing with in flight anomalies is one area which will need to be improved upon, but as computing power increases this does not represent a firm barrier. The ability of the aircraft computers to sort out things such as problematic weather development, engine failures, or traffic congestion will be necessary to a safe flight. Also, the computers will need to instantly compensate for systems that fail, a task many aircraft Flight Management Systems already handle.
Setting up specific airports to accept autonomous aircraft will be the next phase, and implementing Category III type approaches will facilitate smooth landings. Ground surveillance radar now being deployed will provide aircraft with reliable data once on the ground, allowing them to pull right up to the gate without the direction or intervention of ground personnel. Airports that have low commercial use and are far away from congested airspace will provide an excellent test bed. The route between these airports might be fitted with additional surveillance capabilities, to ensure things don’t go wrong along the way.
Once this technology is demonstrated to be successful, automated cargo flights will be the next logical step. This will allow continued enhancements with minimal human risk, and bring any further technical challenges to light before passenger service begins. During this stage of development, a fallback option will allow a remote pilot to take control of an aircraft if its onboard autopilots fail. This would involve a pilot in a remote data center monitoring many flights at a time, and taking over control only if the situation calls for it. The U.S. military is already testing the benefits of using one pilot to oversee the flight of multiple UAV’s in an effort to keep up with the increased demand for UAV operations.7
Another technology that could be a valuable stepping stone on the road to fully autonomous commercial aircraft, would be automating just the cruise phase, and allowing a remote pilot to handle take-offs and landings. This might help the public adapt to not having a pilot on board, while at the same time allowing a better allocation of pilot expertise and therefore improving safety. The pilot would not have the boring cruise portion of flight, nor would he spend hours waiting in terminals and on long commutes. He would merely show up to the data center, remotely perform take-offs and landings all day, and then go home while the mundane portions of each flight are flown autonomously.
After cargo flights demonstrate the capabilities of autonomous aircraft, limited passenger service can begin. A good application for some of the first passenger carrying autonomous aircraft would be international flights over the oceans. Pilots already do an extraordinarily small amount of flying on these vast legs, which would make the transition relatively easy. These flights will provide a good operational test of the computers abilities to deal with unforeseen circumstances, as there is no method of intervention hundreds of miles out over open water. Another reason these flights would be good for testing is because there is not much than can be done to save an aircraft that has a catastrophic failure out there anyway, negating the view of some skeptics that a pilot could have done more had he been onboard.
If the safety record of the early autonomous aircraft lives up to expectations, these aircraft will rapidly expand to other airports as fast as the airports can complete updates to accept them. Small and midsized jets will see great benefits of automation, as they will allow rapid transit to destinations where bigger jets cannot go. Once the foundation is established, there is no limit on the types of aircraft that will be able to take advantage of the new airspace system. General aviation pilots could also install hardware allowing their aircraft to navigate automatically between destinations, allowing them the freedom of sitting back and enjoying the ride. Manual control will still be available and widely used for pleasure and recreational flying, as well as in airshows and races.
Because of the high reliability of modern jetliners, the biggest risk for aviation safety remains the most unpredictable element: the human pilot. Some may argue that the pilot can be better trained or gain more experience, but the fact remains that despite all efforts otherwise; the human cannot perform with the same predictability, reliability, and precision as our computer counterparts. We have seen at how we can remove the pilot from the equation through AI and automation, but it will still be humans building the systems. We must keep this in mind because the machines will only perform up to the specifications which we design into them. There will be disasters, as there are anytime mankind pushes the envelope of what we can achieve. We will need to move past the accidents, learn as much as we can from them, and use that knowledge to improve our systems. As Thomas Edison said in 1903, human progress is “one percent inspiration, and ninety-nine percent perspiration.”8 As we continue to automate the perspiration phase, aviation will reach new levels of safety and efficiency to meet the increasing demands of the flying public.
References:
1 Kaku, Michio. Physics of the Future: How Science Will Shape Human Destiny and Our Daily Lives by the Year 2100. New York: Random House, 2011. Print.
2 Kebabjian, R. (2012, January 4). cause. In planecrashinfo.com. Retrieved December 1, 2011
3 Coniff, R. (2011, June). Drones are Ready for Takeoff. Smithsonian, 37(2), 34-38.
4 Kebabjian, R. McDuffie, P. (Narrator). (2009). Drone Crash Cause for Concern? [Online video]. online: youtube.com. Retrieved December 14, 2011, from http://www.youtube.com/watch?v=yda_vx7H54g
5 Manning, S.D., Rash, C.E., LeDuc, P.A, Noback, & McKeon, J. (2004). The role of human causal factors in U.S. Army unmanned aerial vehicle accidents. U.S. Army Aeromedical Research Laboratory Report #2004-11
6 Kebabjian, R. (2012, January 4). rates. In planecrashinfo.com. Retrieved January 1, 2012
7 Shalal-Esa, A. (2011, December 23). Future drone pilots may fly four warplanes at once. In Reuters.com. Retrieved December 12, 2011
8 Deutsch, David. The beginning of infinity: explanations that transform the world. New York: The Penguin Group, 2011. Print
