mapping software: calculating the time until arrival

Factual Questions
1

I recently borrowed a friend’s GPS (GB: satnav) for my car, just to see how cool it is, and what errors it makes, stuff like that. (It was a Garmin StreetPilot c580.)

I found that if my route was entirely (or almost entirely) on highways, without traffic, and I drove at the speed limit, then the Estimated Time To Arrival was quite accurate. However, if the route was on a street which had stop signs and/or traffic lights, then even though there was no traffic, and I was driving at the speed limit between stops, it took noticably longer to arrive than the GPS expected.

Apparently, the GPS bases its calculations on the theoretical speed limit of each road, ignoring the actual maximum speed that one might drive at. (This was WITH the GPS configured to consider traffic conditions, by the way.)

My General Questions are these: Has anyone else noticed this? Is it common to other GPS systems? Are there any systems which ARE aware of traffic lights and stop signs? How would that feature be referred to in the advertising specifications?

2

I am not familiar with that unit but I am a software person. That is an awful lot to ask of a GPS system (today). A truly accurate system wouldn’t just need to know about traffic signals and stop signs. That takes a lot of data but may be doable. The big problem would be traffic broken down not just by road but also time of day and preferably up to the minute updates. I have no doubt that will happen in the next few years but it isn’t there yet. That is the only way to make the system have the real-world accuracy that you seem to want in a GPS unit.

There is an experiment that you can try for a comparison. Before you do your next driving test, go to www.mapquest.com and plot your route there. Mapquest tends to be pretty accurate for me in their time calculations for Boston routes with average amounts of traffic. Compare the two estimates and see if they are very far off. I would guess that Mapquest is the reference estimate because it has a lot more processing power and current updates than a portable GPS unit. If the two estimates are very close, I doubt you can find anything much better.

3

I too am a software person, and I do realize that if we ask the system to know which traffic lights have twenty cars backed up, that’s an awful lot to ask. We’d need a ridiculous number of sensors and reporting devices for the system to know this sort of information for every neighborhood.

Good thing that’s not what I was asking for.

My point is that if someone went to the effort of accumulating data on which streets are one-way, and which intersections are “no left turn”, and stuff like that, then it should not be much of a stretch to know which intersections have traffic lights and stop signs.

Even without knowing how long the light is programmed to be red for, the algorithm could still use an arbitrary figure (10 or 20 seconds or whatever) and the calculation would be much more accurate than it is now. Even for drivers who do not actually stop for a stop sign, they still decelerate down to 2-5 MPH or so before crossing the street, and if this happens several times, an AWFUL lot of time is lost as compared to simply staying at the speed limit for the whole trip.

Is there any system that does this?

(I’ve used MapQuest a lot, but your procedure doesn’t really solve the problem. There’s no way to know WHY it took me longer than MapQuest predicted. But when I use the GPS, I drive for 30 minutes on the highway without the ETA changing a tiny bit, except for the five blocks in the middle of the trip that were on local streets.)

4

[QUOTE=Keeve]
I too am a software person, and I do realize that if we ask the system to know which traffic lights have twenty cars backed up, that’s an awful lot to ask. We’d need a ridiculous number of sensors and reporting devices for the system to know this sort of information for every neighborhood.

Good thing that’s not what I was asking for.

My point is that if someone went to the effort of accumulating data on which streets are one-way, and which intersections are “no left turn”, and stuff like that, then it should not be much of a stretch to know which intersections have traffic lights and stop signs.

Even without knowing how long the light is programmed to be red for, the algorithm could still use an arbitrary figure (10 or 20 seconds or whatever) and the calculation would be much more accurate than it is now. Even for drivers who do not actually stop for a stop sign, they still decelerate down to 2-5 MPH or so before crossing the street, and if this happens several times, an AWFUL lot of time is lost as compared to simply staying at the speed limit for the whole trip.

Is there any system that does this?

(I’ve used MapQuest a lot, but your procedure doesn’t really solve the problem. There’s no way to know WHY it took me longer than MapQuest predicted. But when I use the GPS, I drive for 30 minutes on the highway without the ETA changing a tiny bit, except for the five blocks in the middle of the trip that were on local streets.)
[/QUOTE]

I work with digital map data, so I can tell you that in general, they carry information about the road itself, so as to enable one to calculate an average speed (and in the near future, a somewhat more accurate calculation will be possible based on average speed when you will be actually travelling that road).

However, they do not yet carry information about every location where there is a stop sign or a traffic light. Various different GPS devices try to compensate for this lack of data by other means. One common one is by adding extra time whenever the path is on a small road that crosses a large road, figuring there must be a traffic light or a long stop sign there.

GPS manufacturers, and other users of digital map data (think Yahoo and Google, for example) are always thinking of new things it would nice if the data contained, so they are always after the two main data suppliers, Navteq and Tele Atlas, to get the data. However, collecting map data is a long, expensive process, so there always has to be a good business case for any new piece of data.

To answer your question partially, oneways and no-left-turns MUST be collected. Without them the data is useless. Traffic lights, at this time, are only nice-to-have.

So it’s not an issue of what GPS system has this, it’s an issue of what map data has this. None does yet, as far as I know.

Ed

5

[QUOTE=suranyi]
I work with digital map data, so I can tell you that in general, they carry information about the road itself, so as to enable one to calculate an average speed (and in the near future, a somewhat more accurate calculation will be possible based on average speed when you will be actually travelling that road).

However, they do not yet carry information about every location where there is a stop sign or a traffic light. Various different GPS devices try to compensate for this lack of data by other means. One common one is by adding extra time whenever the path is on a small road that crosses a large road, figuring there must be a traffic light or a long stop sign there.

GPS manufacturers, and other users of digital map data (think Yahoo and Google, for example) are always thinking of new things it would nice if the data contained, so they are always after the two main data suppliers, Navteq and Tele Atlas, to get the data. However, collecting map data is a long, expensive process, so there always has to be a good business case for any new piece of data.

To answer your question partially, oneways and no-left-turns MUST be collected. Without them the data is useless. Traffic lights, at this time, are only nice-to-have.

So it’s not an issue of what GPS system has this, it’s an issue of what map data has this. None does yet, as far as I know.

Ed
[/QUOTE]

(Missed the cutoff.)

And one more thing: While not so much an issue for the Internet companies, the GPS makers are always worried about data size. Adding information about the location of every traffic light and stop sign would surely increase the size of the data, and perhaps it’s not a tradeoff they’re willing to make.

Ed

6

Thanks!

7

There are far more important things for SatNav manufacturers to deal with before computing average road speed.

Like fitting SatNavs with a Articulated Lorry setting that stops truckies trying to destroy rural villages by taking the shortest route between two points, or waking people up driving through villages that should not have heavy traffic, or getting plain stuck on a non-truck road.

Si

8

[QUOTE=si_blakely]
There are far more important things for SatNav manufacturers to deal with before computing average road speed.

Like fitting SatNavs with a Articulated Lorry setting that stops truckies trying to destroy rural villages by taking the shortest route between two points, or waking people up driving through villages that should not have heavy traffic, or getting plain stuck on a non-truck road.

Si
[/QUOTE]

I can’t get into too much detail, but that kind of data IS being collected by the map data providers, and depending on the navigation system either can be used now or will be able to be used soon. But only by more expensive GPS systems intended for the trucking market.

Ed

9

[QUOTE=suranyi]
I can’t get into too much detail, but that kind of data IS being collected by the map data providers, and depending on the navigation system either can be used now or will be able to be used soon. But only by more expensive GPS systems intended for the trucking market.
[/QUOTE]
The timescales I was hearing was 18 months or so. Of course, some Lithuanian truckdriver making has way across Europe will probably still buy a retail Navman on the cheap. The EU needs to legislate on this one.

Si

10

Interestingly enough, I had somewhat the opposite experience with my Garmin eTrex. I put maybe 1,000 miles on it before I found the ETA function, which I used for the first time on the return leg of a road trip. When I turned it on, our projected ETA was roughly 2 hours (it actually gave us a clock number, as in 10:45 PM) During those 2 hours the ETA didn’t waver one bit, and we got home exactly one minute later than our projected time. But the kicker was: we were driving well over the speed limit, so I can only draw the conclusion that the GPS had learned how fast we drove and took that into account when making its estimate.

11

[QUOTE=suranyi]
I can’t get into too much detail, but that kind of data IS being collected by the map data providers, and depending on the navigation system either can be used now or will be able to be used soon. But only by more expensive GPS systems intended for the trucking market.
[/quOTE]
The Garmin StreetPilot c580 that I was using has several settings, including Car Truck Motorcycle Ambulance and at least two others. I can’t imagine what that setting would affect other than which roads you’re allowed/disallowed on.

12

[QUOTE=Keeve]
The Garmin StreetPilot c580 that I was using has several settings, including Car Truck Motorcycle Ambulance and at least two others. I can’t imagine what that setting would affect other than which roads you’re allowed/disallowed on.
[/QUOTE]

The new units for trucks allow, or will allow, a much more precise selection than that. They will allow you to specify, for instance, the height and weight of your truck, and whether you are carrying certain kinds of hazardous materials. The road data being collected is precise enough to distinguish between all these different allowed characteristics.

Ed

13

I had a Garmin Street Pilot III and now have a Garmin nuvi 550. They both behave the same way. I have studied this somewhat. It assumes I will go faster than the speed limit, especially on interstates. If you go 55 MPH you will arrive later than the projected time. It also completely ignores the effects of traffic control (stop signs and red lights) on local roads.

On MapSource, the desktop companion software for it, there is a set of options to set driving speeds, but it is only for the software and does not download to the unit (where there is no such option). Here are the defaults in MPH:

67 Interstate highways
58 Major highways
45 Other highways
35 Collector roads
25 Residential streets

I took my Street Pilot III on a trip to the southwest last spring and noticed that it grossly overestimated travel time on one particular interstate in Utah where the speed limit was 75 MPH. I assumed that was because either the speed limit had been raised since the map data was generated, or the unit uniformly assumed a speed of about 67 on any interstate regardless of actual speed limit.