Drones CAN'T deliver goods ... oh. Air traffic control system backed by NASA, you say? Recently we wrote about how developing city delivery drones was a fool's errand. Yet Startup Airwave is working with NASA to develop an air traffic control system for drones. Who’s foolish now? Airware was founded in May 2001 by Jonathan Downey, an electrical engineering and computer science graduate from MIT, and is based in …
"Every flying vehicle in that environment is controlled by the system. That's the only way possible, we think, to rule out controlled vehicles crashing into un-controlled ones."
That's the most worrying thing I've heard today. What happens when a vehicle develops a fault? I guess others will crash into it because it moves in a way they don't expect. What happens if a human-controlled aircraft comes into the area? Crashes get far more serious. It also implies no situational-awareness, so if a tree grows they may well fly straight into it.
No doubt this sort of thing has been considered, but the way they present it does make me concerned...
Given that there will be birds flying in the area I'd imagine that certified drones will have to have sufficient autonomous collision avoidance sensors on them to cope sufficient with most negative scenarios. Pretty much like autonomous cars will have to cope with kids stepping out in front of them.
I've been skeptical for a while now on the viability of drone delivery platforms, but lets *assume* they become viable in the near future (at least technically speaking) then the question becomes *what next*
There is no way multiple providers of such platforms (be it commercial companies, organizations or anything else) can have drones flying around simultaneously - not without them either crashing into each other or (if they are smart enough not to bump into other drones) hover around competing for airspace to get to their destination. Queue chaos for the former and congestion for the latter.
So then, i can think of 3 solutions for this problem:
1. either these companies start to develop some sort algorithm to allow drones from different manufacturers to talk to each other mid-flight to sort out mentioned problems (is anyone developing this now?)
2. there is some sort of airport-control-centre like setup in cities (too expensive)
3. or, more logically, a single company takes up control of all drones from all manufacturers through some sort of unified interface or communications protocol.
Any one of these solutions would take some good and hefty development to achieve, i'm glad NASA is enlisting here.
Still skeptical on how achievable a fleet of drones is, at least for the near future
.
I think the point of the research with NASA is to develop option 3 as a national ATC system for drones.
Obviously if/when an ATC system is implemented, all drones will have to be FAA (for USA) certified to operate correctly within the ATC system.
This will suddenly confront the drone makers with a lot, but not all, of the regulation and cost of manned aircraft, so drones will be less cost effective, but probably still worthwhile.
It's the same as manned flight; back in 1900 you didn't have to get your manned flying machine FAA certified and your pilots qualified to work within the air traffic control environment.
Avoidance is a 3d issue. 2 devices can be at the same X,Y coordinates at time T, as long as coordinate Z is different.
The aircraft standard is that different levels are (supposed to be) used for different flight directions, normally at 500 foot levels. This means you only have to look for stuff crossing your path, instead of closing at high speed.
There are also more prosaic rules, like giving way to the right and conventions about going down/up if paths are crossing, dependent on which direction one is travelling in.
Humans are bad at dealing with more than 3-4 objects moving around at once, as anyone playing Galaxians will attest. Machines should be able to do much better, more quickly, because they don't get distracted and because they should be able to chatter amongst themselves both to agree on directions and warn of uncontrolled objects (such as those pesky birds)
[Rant]
UAV's have a lot of benefits, there is a lot of things they can do smarter and cheaper than anything else (reconnaisance, mapping, etc.), but acting as a delivery service is never going to be anywhere near as cheap or effective as regular delivery from a van.
First, there is carrying capacity - a drone cannot carry a great deal of weight; second, there is fuel economy - a drone uses significantly more fuel per km per kg then any automobile on the planet; third, there are all of the other issues associated - securtiy, regulation, the actual delivery, etc. Your Fed Ex package might occasionally get thrown over the back fence, but in general if they cant deliver it, they take it back or maybe they give it to a neighbour, or it goes in your letter box, but a drone is unlikely to be able to do any of that. So your parcel will get dropped on the lawn and you can kiss that goodbye in many neighbourhoods.
So bring on more practical uses of UAV's but can please stop the nonsense about drone deliveries, it distracts from the useful applications!
[/ Rant]
I can't see them being a reality any time soon, but I am prepared to think that they could be. In relation to your points...
No one said they have to replace van/truck/train delivery, they would complement it rather than replace it, at least in my lifetime.
As the delivery would be in the near future (expected within the hour I would guess to make the drones worthwhile), then you would be at home, it's just been ordered. You may have to put out a landing or drop marker for it as well. There would be no point ordering by drone if your delivery option is a week next Tuesday.
As for fuel, they would be designed for close area flights so a drone could work out more fuel efficient than a van in congested streets.
If you can ever envisage a self driving car doing deliveries then it would surely be a lot easier to use a UAV with far more space and less obstacles?
People seem to be stuck with quad copters and home deliveries from Amazon in their heads; the article talks about UAV's in general, so that could be anything, from (yes) quad copters, to air ships and even quite hefty aircraft. Whatever UAV platform people come up with, it is essential to have a traffic control system in place first!
And, frankly, all of you are thinking too small anyway. Maybe I want myself to be delivered to a destination as the crow flies (for some this may cut a great deal of congestion/fuel usage). I mean, how else am I going to get the flying car I was promised in the future? I'm not going to pass a pilots licence any time soon.
Well screw you
I have a perfectly adequate back garden that could be used for deliveries along with thousands of others!
I guess we better scrap the whole idea because it doesn't benefit you.
Fuel efficiency is not bad when that said van is stuck on the motorway for 4 hours stopping and starting in traffic etc etc, not saying get rid of the van but a big old van but supplement it.
@AC regarding Pollution, Its a presumption as i cannot tell the future, but im going to guess that they will be electric powered, not petrol.
so resolving the fumes but the amount of buzzing would get irritating, I would reckon that to be the bigger issue
•The drone's aerial environment is 3D-mapped to a high degree of granularity: similar, perhaps to the roads mapped by Google's self-driving cars.
Problem being the road environment is much less prone to change than the low-level aerial one. Even things like trees growing, sprouting leaves and blowing in the wind for example would be much more variable, not to mention similar temporary hazards like flags/banners, washing lines and other nice drone-snares.
•Every flying vehicle in that environment is controlled by the system. That's the only way possible, we think, to rule out controlled vehicles crashing into un-controlled ones.
Even leaving aside the uncontrolled elements like birds, quite how would control be achieved? Everyone knows quite how difficult it is to get a signal into built-up areas (sat navs vs urban canyons for example), plus the tussling between the requirement for overall control vs local autonomy so any drone could avoid the uncontrolled obstacles already mentioned.
If this kind of automated air traffic control was so easy, we'd already have it for our existing transportation like cars (ok, not air traffic control, but general control) and more integrated with that for normal planes or even small ones.
Plus to carry anything worthwhile (ie more than just a single small package over a very short distance) you'd need a large and powerful drone which would probably be getting to the realms of a small helicopter equivalent anyway.
To be fair - nowhere in the article says it will be easy. They are just saying that there needs to be one, and these chaps are going to have a crack at it. If the system cannot cope with the issues you mention then I would suspect that it won't be very successful.
Proper drones have a lot of collision avoidance etc stuff in them anyway, using SAR/LIDAR/RADAR as part of an array of sensors to give as much situational awareness as possible. So this would deal more with the routing aspects - in the same way that ATC doesn't fly a 747 nowadays, that's the captain's responsibility.
In summary, air pilots and captains can talk to the meatbags at the ATC via radio links to convey the relevant info to avoid collisions with other aircraft, get wind data etc. This is looking at an M2M version of your traditional ATC to deal with the fact that the pilot is a processor, not a person. It also doesn't rule out human operators for the ground based routing system.
Personally I'm surprised that this is the only company we've heard of trying to do this, as the UAS sector will be screaming for this capability in ~5 years time.
@ a twig
The statement about "proper drones" having SAR/LIDAR/RADAR to give them as much situational awareness as possible is grossly misleading.
There is a night and day difference between putting these kinds of sensors on a fixed wing airframe, like the Predator, Global Hawk, or a South African Hunter II, and putting them on a comparably-sized rotorcraft. Even with the most sophisticated military sensors and humans intimately in the loop exercising direct system control errors still occur at a frequency that will be categorically unacceptable in a domestic civilian environment.
As for the kinds of small "delivery drones" being proposed: As others have noted, the utility of the technology for applications being touted by proponents of the technology is already questionable because of the limited payload capacity. Adding sophisticated RF sensors is not a viable solution.
People who do not actually work in the field vastly underestimate the challenges of machine perception. Again, I reference Tom Fawcett's work in ROC curves as a readable explanation of the science involved.
As for the time scales projected, the automotive sector provides a good case study.
http://en.wikipedia.org/wiki/Autonomous_car
The 1980's EU/Mercedes (albeit with human observers on board) successfully ran under fully-autonomous control on the Autobahn, if memory serves, at reasonable highway speeds. For reasons well-documented in the Wikipedia article, we are still a decades away seeing numbers of fully-autonomous self-driving cars except on specially configured highways. .
@ Hargrove
I have worked briefly in the UAS field and totally agree with your points. I don't think a Quadcopter delivering something to your back garden will be a viable proposition for a long time if ever. However the issues you raise aren't directly what I was talking about, as these refer to issues within the airframe, not the ATC.
In context:
"Proper drones have a lot of collision avoidance etc stuff in them anyway, using SAR/LIDAR/RADAR as part of an array of sensors to give as much situational awareness as possible. So this would deal more with the routing aspects - in the same way that ATC doesn't fly a 747 nowadays, that's the captain's responsibility."
I was merely pointing out that the scope of this UAS ATC system would not extend to collision avoidance with pigeons etc, in the same way that ATC doesn't now. All it will deal with is routing for all UAS in the area. It is up to the individual UAS to be able to fly competently.
Essentially being able to fly a UAS without hitting things is the manufacturers problem, not Airwave's...
For the better part of a half century, we have had the technology for autonomous navigation in constrained environments. From a system design and integration perspective it does not get much constrained than ground rail transport. Yet after fifty years unmanned systems for surface transport are used only in limited situations, where the entire system and its operating environment are owned, and controlled by a specific central authority--airport people movers, warehousing operations (where the "rail" can be a painted line on the floor") and the like.
Where human life and safety are at stale, humans monitor and exercise ultimate control. Where interoperability with humans or other entities capable of independent movement is involved (for example, in streetcar systems) those human operators are on-board the vehicles themselves.
Extending this to effective operations in airspace will require solutions to problems that those pushing this technology are not even considering. Anonymous Custard sites some examples. (His comments also recall a previous technological prediction about the use of helicopters for personal use that failed came to nothing.)
In addition, the presumptions being made about the effectiveness of on-board autonomous sensors are fatally-flawed. Again, I recommend Tom Fawcett's works on ROC Curves. The bottom line is that there is no such thing as a free lunch when it comes to the problem of true vs. false detections of objects of interest in an operating environment.
It's a bird. . . , No wait, it's a plane. . . no wait, it's another UAV . . . No. . . It's a flock of pigeons. . . "WHAT'S THAT ON THE LENS? FLYING BLIND MUST LAND NOW! Wait. . . was that a school bus windshield I just went through?
"Problem being the road environment is much less prone to change than the low-level aerial one. "
You didn't honestly write that did you?
Every single journey to work the road environment for me has changed. From pedestrians crossing, to road works, accidents, traffic signs, obstructions, delivery vehicles etc.
The challenge for an autonomous car verses a few trees that sway in the wind or a flag pole (which the uav will be flying way above any way) is massive.
@A_C a couple of posts above - OK, maybe adding the word "random" into my line above that you quote would have been more accurate to what I meant.
At ground level, everything is in 2D, and a lot of the changes (crossings, roadworks, parked vehicles and perhaps even accidents) would be fairly easy to have notified and/or detected (crossings and road works would presumably have some sort of transmitters to make all the automated stuff aware there they're there, and vehicles will have that to talk to one another anyway for synchonisation and ensuring that their autonomy works).
In the air, you have another dimension to worry about (although in fairness also another one to maneuver in, presuming you have some height), but the random variability either tends to be more natural in origin, or not related at all to the transit system and so less likely to be squawking out that it's there and so will need to be dynamically reacted to.
But in any case I would guess the economies of scale are working against them too. The drone use is the pizza delivery model, whereas the requirement is more the UPS/Fedex multi-drop model to make it worthwhile. I entirely agree that if the use of drones is going to have a future then something like the ATC will be needed and will be fine for the "highway" part of the journey, but it's the local bit (especially the "parking" at the end) where things will fall down.
Better. On the local bit, there already exist precision-approach radars to handle the situation where neither ATC nor the aircraft have even IFR minimums. Scale-down wouldn't at all difficult. Similarly, there are automated landing systems for shipboard aircraft recovery. Pilots being pilots, you practically have to hold them at gunpoint to let the machine do it. Landing accuracy to a (very) few inches. The technologies are more than four decades old and knowing them intimately, quite amenable to cheap implementation. Barring the patent thickets.
Noted to self. Probably more fun than an autonomous ground drone.
Birds seem to have been managing to avoid other birds for quite some years now, given that drones are likely to be relatively slow and not have a bloody great jet intake, I think birds will manage to avoid most UAVs.
The current time scale mentioned for development of the system is around 5 years if batteries have not improved enough by then, I guess there will just be delay but it is beginning to look as though the idea is here to stay.
Great for delivering batches of spidery things to assist the cops/court servers/debt collectors and Amazon. Not to mention pizza and other take aways, although that will spoil a classic porno story line with a drone joining in with three gals and a guy.
Vancouver has been running automated trains for over 25 years, though there is a manned central control room. Default for some issues is full system shutdown, happened a couple of times this year it seems. Safety comes at a price in this case you can be stuck because someone decided to force doors and go for a walk when their train stopped for 5 mins
It's that sort of lateral thinking that I see as the way forward.
Too much of this discussion is black and white Van v Drone where the aerial vehicle is (justifiably) seen as unable to replace the Van in it's current use.
UAVs are not Vans and trying to replace one with the other is too simple a viewpoint.
The UAV bit is only part of the delivery chain, maybe lorries will pass through local set off points and automated systems pick and dispatch from motorway service stations.
The flight corridor should not be insurmountable if the various legs of the system are tailored to the transport vehicle's strength, move three tonne on direct M-way route, that will be a truck, 150g medicine across three miles of unclassified roads, UAV, but ONLY for the air-efficient leg.
Have the traffic control and some pathways from mobile phone masts, have mesh corridors and crash zones so that if the system goes down the UAV's head to farmer Giles back meadow (the one he's paid UAVSA for) or the local sewage treatment works, better to recover the bits from some sh!t than hit a school.
It's coming and many can see it evolving, others say there will always be vans just like messenger boys.
Mr Downey disses open source projects, probably because he can't make any money on them. Arduino based projects like Multiwii and Megapirate have come a long way in a short time. Most hobbyists see that a drone delivering stuff using present motor & battery technology as a pipe dream.
I build sub 200g drones for remote video, with GPS and limited sensors. If these had to avoid another drone then a >10Mw transmitter & receiver could be used to determine if another drone was in the immediate area and they could handshake and update there track to avoid each other. All without Sky Net (sorry Airware). Hitting trees and birds is a different matter & requires work as current sensors have a max range 50 foot or so.
Go back to cadging cash of the military Mr Downey.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
