Who is a Better Pilot: a Computer or a Human?
A vastly expanding civilian Unmanned Aerial Vehicle (UAV) market is leading to the day when UAVs must be allowed to operate freely in the National Airspace System (NAS). The main obstacle preventing this from happening is the distrust of the ability for a UAV to avoid collisions with manned aircraft. There has been a vast amount of effort put into researching UAV Detect Sense and Avoid (DSA) systems with the hope that an Equivalent Level of Safety (ELOS) can be achieved for unmanned compared to manned flight. The current debate is what type of DSA will provide the required level of safety to prevent increasing the hazard to manned aircraft operations.
There are many different types of DSA systems currently undergoing research, from visual cameras interpreted by humans to automatic detection systems incorporating lasers. ASTM International, an organization that develops industry consensus standards such as those that govern Light Sport Aircraft design, has released a standard (F2411-07) on the requirements for an Airborne DSA system that specifies the minimum capabilities of a DSA. However, this ASTM standard does not (and should not) mandate what type of system must be used.
The overriding question regarding what type of DSA will eventually be used is whether or not the pilot operating the vehicle should be “in the loop”. Can a UAV’s computer detect an aircraft and maneuver to avoid a collision automatically as effectively as a pilot on the ground could view a high resolution video screen and perform the same operation? There are many factors that must be considered before this question will be appropriately answered, such as pilot/computer response times and communication latency. However, a human has the ability to make decisions based on variables that may never have been identified as pertinent by a computer software engineer. Thus, a human interaction with the DSA system may not be the most “precise” option, but probably the most dynamic. Will we ever see 100% autonomous UAV operations in the NAS? Only time will tell.
8 Responses
to “Who is a Better Pilot: a Computer or a Human?”
3 Trackback(s)
- Jul 16, 2010: Mark Blanks
- Jul 16, 2010: Mark Blanks
- Jul 16, 2010: markblanks
Well coming from the unmanned side of things as I do DSA is big on out minds, at a symposium last week at RAF Waddington the thought of DSA systems being better than manned ability came up and then you have to take that to the point of…. its not fair if the manned stuff is not as clever
I think you have identified two key issues that, presently, remain largely unanswered:
1. Computers fly better than pilots within a given parameter set
2. Humans respond and fly better than computers “outside the box.”
Reconciling, synergizing, these two areas is the challenge for us, the aviation leaders of the future.
Mark
Well the ? is who is the better pilot. Since a pilot is not only required to fly but also make changes while in flight. There is no way a computer can do anything if it is not programed to do so. Ask any pilot if during a flight how many time he or she has turning off the autopilot or at least put it is a heading mode to make a change. If a computer was at the control and a cloud is ahead that is still in the building stage and not showing on the radar it might fly through it and could give passengers or cargo a rough ride or even stress the aircraft. I believe the human pilot is the safest way, with the help of another human to keep human factor mistake to a minimum. Let the computer fly and let the humans do the “piloting”. There is plenty of room in the Flight Deck:)
Marc makes a good point about the difference between “piloting” and “flying”. However, I, personally, don’t want to loiter at 60,000 feet over the oil spill letting the computer fly me in circles while it takes pictures.
Marc,
The problem is that DSA systems are getting so advanced that the number of “possible” scenarios where they will not work is being reduced to almost negligble. There are systems out there that would “see” the building cloud that you are talking about and be able to take evasive action. Also, most UAV applications currently under research are not meant for carrying people or cargo. Small UAVs that house sensors and cameras that will not complain about a rough ride are the most likely first applications that we will see for unmanned systems.
If we were to try to prove that a UAV can provide an ELOS to manned aircraft, then we would need to look at how safe current manned operations are. Currently, in the VFR environment, pilots get distracted and are not always vigilant about scanning for other aircraft. I would say that it is very likely that a DSA system would provide fewer “close calls” than a manned aircraft over a long period of flight time. However, I do feel that a human pilot should be in the loop because there are situations that have not even been considered yet that may arise and a computer, even with advanced artificial intelligence, might not be able to respond appropriately.
Mr. Mortimer could probably provide us with some details on how DSA systems are better than manned ability. I know the technology is amazing, but I cannot forsee an automated aircraft as being versatile enough to operate in the U.S. NAS where there is heavy air traffic volume ranging from commercial airliners to hang gliders. I am not an “expert” on UAVs, though, so my opinion may be completely inaccurate.