The Naked Scientists

The Naked Scientists Forum

Author Topic: Driverless BMW  (Read 4747 times)

Offline DoctorBeaver

  • Naked Science Forum GOD!
  • *******
  • Posts: 12656
  • Thanked: 3 times
  • A stitch in time would have confused Einstein.
    • View Profile
Driverless BMW
« on: 04/02/2008 08:01:04 »
I was watching an episode of Top Gear a while ago and they had a BMW 3 series on there that those clever Bundesgenii had fiddled with. They drove it round the Top Gear test track and it mapped the lap using GPS and other clever gizmos.

Then, with Clarkson on board in case anything went wrong, it went round the track flat out with no-one driving. I think Clarkson needed a change of underwear afterwards.

But, aside from the humorous aspect of the terrified Chipping Nortoner making a mess on the seat, I was very impressed with the application of technology.

Now there are adverts for a Volvo that applies the brakes for you if it thinks you're going to hit something. There's a Merc that does the same thing. The new Lexus 600i can park itself.

Are the days of the car driver numbered? How far are we from entering our destination into our GPS devices and letting the car do the rest?


 

another_someone

  • Guest
Driverless BMW
« Reply #1 on: 04/02/2008 12:58:44 »
I would think we are probably about 15 to 20 years from that.

We have only just last year, or the year before (cannot recall) had someone will the million dollar prize (or some such) from DARPA to get a car/truck to run around a previously unseen track, with obstacles and other vehicle, without crashing into anything (most of the other contenders managed to crash into something).  Doing it once, and doing it every time, is another thing - and it wont take many confused automatic vehicles having fatal collisions (even if 50% of them don't) to start the law suits going.

Following a known track on GPS, with few obstacles, is one thing; having to watch out for pedestrians about to step off zebra crossings, children running out between cars, road works, 30mph speed limit signs that weren't there yesterday, police officers flagging you off the road, etc., are all far more complicated than racing around an empty tract on GPS.

That having been said, there are beginning to creep in what are known as cybercars.  These are vehicles that do as the BMW did, for public use, but not for public roads, so you can design a road infrastructure that manages the limitations of a blind motor car (e.g. they are presently limited to 20mph, and if there are any strange diversions, etc., they can be reprogrammed in some way rather than expecting machine vision to detect this in real time).

I suspect there will be a convergence between the blind cybercars we presently have, and the part time driverless vehicles (e.g. self parking vehicles, once the driver has told the vehicle this is where to park, and the driver remains responsible for any police officer who tells him to stop the vehicle and get out).

Ofcourse, one thing I would be interested to know is how, if we do develop fully driverless vehicles, we deal with the potential use as terrorist guided missiles (after all, it is perfectly reasonable, if I have a parcel to send somewhere, and I have a driverless vehicle, it could just as easily be a passengerless vehicle, containing only the parcel that needs to be delivered - but who is to say what is in that package, and whether it will explode upon its arrival at its destination?).
 

Offline engrByDayPianstByNight

  • Sr. Member
  • ****
  • Posts: 105
    • View Profile
Driverless BMW
« Reply #2 on: 06/02/2008 05:22:20 »
Cornell university, where I'm currently doing my Ph.D., was one of the six teams that made it through the end in the DARPA unmanned vehicle contest (although it wasn't the winner). The team had an open-house day a couple months ago with the vehicle in display, and I got to see it in close range.

Although I didn't follow that contest and don't know much about the technology being used in today's and future cars, my personal take is that the GPS as the only guide to navigate an unmanned vehicle across long distances with various terrains is probably not a reliable way. GPS signal can be easily disrupted with bad weather or when the vehicle is in a tunnel (i.e., not exposed to open sky), which could happen fairly frequently. Serious malfunction of the GPS satellites could render these cars inoperable.

A practical, but more expensive approach, to realize true driver-less vehicles, in my opinion, would be the deployment of extensive ground infrastructure. Equipment that is analogous to today's wireless cellular networks (base stations, executive cellular processors (ECP), etc.) would need to be implemented along highways, in the woods, in the deserts... that would serve as local guides for a vehicle in vicinity. The advantage of this approach is that the information the vehicle receives pertaining to local traffic conditions would be more accurate and up-to-date. One main disadvantage is of course the cost of such a large-scale implementation of the infrastructure. So in essence, from this perspective, realizing driver-less vehicles is a distributed-network problem. But of course, GPS will still have an important role to play in this model.

It would take probably more than 10 years before we see a successful prototype of a driver-less car in its true sense. However, I think that at least for the time being, we'll be seeing a car that's an integrated hybrid of on-board computer instructions and human interaction, where the computers would be more responsible for making split-second safety decisions (collision avoidance, slowing down when driving on slippery grounds, etc.), and a human driver would be doing the mundane tasks such as keeping the wheels straight and signaling before turning.

As a side note, most commercial aircraft today have auto-pilot systems which allow a plane to fly and land unmanned as long as the conditions are good (by the way, do these auto-pilot systems allow a plane to take off on its own? I'd say no since, well, why would we want to do that if the plane's already on the ground?). My pure guess is they rely on the towers at different airports along the flight path to guide them every step of the way, right? Someone with more knowledge on how aircraft auto-pilot works?



 

Offline DoctorBeaver

  • Naked Science Forum GOD!
  • *******
  • Posts: 12656
  • Thanked: 3 times
  • A stitch in time would have confused Einstein.
    • View Profile
Driverless BMW
« Reply #3 on: 06/02/2008 07:46:02 »
Autopilots are a bit different. You don't often get kids running out in front of you at 30,000'  :D
 

another_someone

  • Guest
Driverless BMW
« Reply #4 on: 06/02/2008 13:00:27 »
Firstly, autopilots can take an aircraft to within a few feet of the ground (maybe even down onto the ground) using MLS.  These days, it is possible to land an aircraft with only a few feet of visibility (but it is the responsibility of the tower to make sure the runway is clear - even in the absence of autopilot, and with good visibility, if something crosses the runway just as the aircraft touches the runway, you will still have a tangled mess of metal).

The DARPA unmanned vehicle contest did use, but only in part, GPS.  GPS is now accepted as a primary tool for air navigation (the old method, using ground based radio beacons, is now I believe largely obsolete - although am not sure if it has been, or will be, phased out - maybe some older aircraft still use it, despite its limitations).  The point is the DARPA test was not just about navigation but about obstacle avoidance in an unseen course, so it included tools such a laser rangefinders to detect obstacles.  Despite, for the first year, there being vehicles that did complete the course, many others were involved in collisions due to a misreading of the obstacles.

GPS, for civilian use, is regarded as reliable (this may be an error - as you say, potentially it can be switched off at any time, although jamming is more difficult, but not impossible - but that is also why the Europeans would like to have their own GPS system, and the Russians already do have their own).  The problem is that GPS is only useful to manage a static environment, not a dynamic one.  Running on GPS is like running on rails, but it does not warn you of obstacles along your path.  This is fine if you are in an environment where you might assume the track is empty (e.g. it is fenced off, as is true now for much of the railway network, and so we do now have driverless trains - the DLR being one); but you still cannot have driverless trams because tram lines are not fenced off.
 

Offline DoctorBeaver

  • Naked Science Forum GOD!
  • *******
  • Posts: 12656
  • Thanked: 3 times
  • A stitch in time would have confused Einstein.
    • View Profile
Driverless BMW
« Reply #5 on: 06/02/2008 14:07:17 »
In 1980 I flew back from Bombay in a 747. After we had touched down at Heathrow the captain announced that due to poor visibility we had been on autopilot ever since the plane crossed the coast at Margate and that the auto pilot had even landed the plane as there was too much fog for the pilot to land us.
 

Offline engrByDayPianstByNight

  • Sr. Member
  • ****
  • Posts: 105
    • View Profile
Driverless BMW
« Reply #6 on: 07/02/2008 01:22:56 »
Doc, that was a pretty fascinating experience no doubt. I could almost see the stunned disbelief on everyone's facte when the captain told the story. If an airplane back in 1980 could do this, I wonder what the more modern aircraft of today with more sophisticated avionics can achieve with auto-pilot, besides handling more challenging weather conditions (like flying in a blizzard)? Greater stability of the aircraft perhaps?

 

another_someone

  • Guest
Driverless BMW
« Reply #7 on: 07/02/2008 02:48:41 »
Doc, that was a pretty fascinating experience no doubt. I could almost see the stunned disbelief on everyone's facte when the captain told the story. If an airplane back in 1980 could do this, I wonder what the more modern aircraft of today with more sophisticated avionics can achieve with auto-pilot, besides handling more challenging weather conditions (like flying in a blizzard)? Greater stability of the aircraft perhaps?

The latest aircraft, such as the Boing 777, airbus 320 and later, are all fly-by-wire; so I would imagine a fair amount of electronic stabilisation is possible (although, at least in the case of the 777, there is a manual fallback, so it clearly does not rely on the electronics for dynamic stability, in the way some military aircraft do).
 

The Naked Scientists Forum

Driverless BMW
« Reply #7 on: 07/02/2008 02:48:41 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums