Fanum
Toubab
Yep.....the software's as updated as the online updater thinks it should be.......all the patches and latest version of mapsource etc 
More routing oddness
- Thread starter Fanum
- Start date
The theory that the cpu is non-deterministic doesn't really hold. If a computer does something different on 2 runs of the same program, it's for a reason encoded in the program or the data it's processing. If computers 'did what they liked' then they are no longer much use. In the computers we sell, if one cpu goes and drives different data on the bus during a cycle, compared to the others, it's rogue, and gets shot before the instruction does any damage.
The decision tree pruning based on time theory is way more likely.
The decision tree pruning based on time theory is way more likely.
PanEuropean
Registered user
ianf said:
It's even more likely in light of the fact that the newest GPSRs (for example, the 27xx and 28xx series, possibly others) will recalculate routes in real time in response to changing street attributes that are based on time of day restrictions. This is why the 27xx and 28xx GPSRs sometimes pop up a message that says "A better route is available, do you want to take it?"
I gotta tell you, I was awefully confused when I first saw the GPSR doing this.
My initial thought was "Well, if there is a better route, why the heck didn't you figure that out the first time around?" It wasn't until I learned that the proposal from the GPSR was a result of time of day restrictions that it all made sense to me.Here's an example of how it works: Let's say you calculate a route from A to B, and begin navigating. Towards the end of this route, there is an intersection that has a 'no turns allowed' restriction in effect from 16:00 to 19:00 daily. You start driving at 15:30, and your estimated time of arrival at the destination is 15:50. The GPSR will calculate a route that has you turning left at this intersection, because this is the shortest or fastest way to get to the destination.
But - you stop halfway to fill up with fuel and buy a coffee and donut. In the meantime, your GPSR is looking at the new ETA, and it sees that because of the unforseen delay at the petrol station, you will be arriving at this turn-restricted intersection at 16:03 - too late to make the turn. It will then recalculate the best way for you to get to your destination, and up pops the question "A better route is available, do you want to take it?"
Michael
ianf said:
I don't claim to be a computer techie, and I'm getting away from the point a bit, but surely it depends on how 'small' you go, ie how deeply you go into it. Fanum has already said that his computer (GPS) did something different on 2 different runs, and now we're all trying to work out why. Sure, there may be reasons why it happened which originate proximately from a programmer's decision, but the fact remains that from his point of view, it happened. Now if you go to a smaller and smaller level, I bet when you get down to bits or even electrons, no-one can predict exactly what's going to happen every time. The program and hardware is designed to minimise or compensate for this random element (which is why it seems to rarely happen), but I believe it must be there. You may say that it's simply a case of having enough technical or programming knowledge to know what happened, but, heisenberg-style, there will come a level where no-one actually knows what it is that happened. And that's the same thing as saying the computer did it itself. I'm afraid this is a bit metaphysical !
At the very least, I am reliably informed that cosmic radiation does occasionally switch the status of a bit.
And you DO admit in your post that you get 'rogue' performers, which you just ditch without a) it having been detected in the factory CPU test and b) never knowing why it gave a rogue result.
PanEuropean
Registered user
gasman said:
Yeah, that may be true, but time of day turn restrictions is simpler for me to get my head around.
Michael
Dahoum
Registered user
gasman said:
You mean there really IS a little man with a load of maps inside my GPS
It's possible but rare for memory to get a single bit error caused by a cosmic ray going through the silicon. Ballpark, I think we used to report on it being detected about 1 time per year per system. That memory kept syndrome data that allowed us to see it, and a partner board we'd just copy it back from.
Now that you have ecc memory it's fixed up by the hw on the board. Doubtful a gps has this sort of memory, and the mapping data is held in flash memory, which might not be quite as vulnerable as s-ram, but a bit flip in the mapping data is going to bugger up a lot of info given it's compressed the map.
As for an instruction executing wrongly, the odds of that happening and it being advantageous are pretty low. Usually it causes a fault. Not sure how they are handled on the cpu in a gpsr, but you'd likely see a reset or halt.
Time to get out there and ride I think.
Now that you have ecc memory it's fixed up by the hw on the board. Doubtful a gps has this sort of memory, and the mapping data is held in flash memory, which might not be quite as vulnerable as s-ram, but a bit flip in the mapping data is going to bugger up a lot of info given it's compressed the map.
As for an instruction executing wrongly, the odds of that happening and it being advantageous are pretty low. Usually it causes a fault. Not sure how they are handled on the cpu in a gpsr, but you'd likely see a reset or halt.
Time to get out there and ride I think.
PanEuropean
Registered user
Yea, really. If cosmic rays can mess up the navigation directions, I'd hate to think of all the problems that speed bumps and potholes are causing (not to mention all those radar and laser rays from the speed cameras).
Michael
Michael
Bezzler
Guest
Fanum said:I set a route to a destination......the Quest gets me there nicely, thanks Betty
Then I hit and hold the 'find' button to get her to take me home.....and she takes me a different way home???
I think my Betty's got a sense of humour, or at least she likes to vary her route home for a bit of interest!
I'm sure there are reasons for it, it doesn't bother me 'cos all her routes are valid, but it doesn't make any sense to me
What do you expect ffs it's female apparently..................
Quick were's the bomb shelter...
Bezzler said:What do you expect ffs it's female apparently..................
Quick were's the bomb shelter...
are you suggesting we could have a GPS with no maps at all, which just asks passersby for directions through it's loudspeaker, with a mirror on top and a pink colour scheme ?
Schwarz Baron
Registered user
I'm outta here, this is heavy shit man!!! I'm away to smoke some weed instead. 
HMR
Registered user
I'm sorry that I have to tell you - No, there is no random element involved!gasman said:
It was not two identical runs. Fanum did not press the recalculate button on exactly the same spot. Neither did he press the button on exactly the same time of the week. There where probably also GPS signal reception differences. This is enough to explain the different routes selected.
There are other possible reasons. Software bugs for example. A read address bug may make the GPS behave as if a random element was involved. Still, a GPS is a digital computer that operates "exact" and not an analog instrument with an environment dependent precision.
PanEuropean
Registered user
HMR said:
Quite true - the GPSR itself has no judgement, cognitive ability, or discretion. But - the huge quantity of road attributes - which change frequently even from block to block along the same street - are intended to give the GPSR route calculation engine the ability to make 'judgements' about which route would be best (shortest / fastest).
So, you could have a situation where you are at one point on 'Main Street' and you get a calculation that tells you to continue to drive along Main Street, and if you go forward a couple of blocks and then ask for a recalculation, the GPSR will send you via a different path that takes you off of Main Street.
We've all seen that the cartography databases are growing in size by about 20 to 25% each release - this is not because there are 20 to 25% more streets being added, or a similar size increase in the number of gas stations and restaurants - it's because so many more attributes are being added to existing stretches of roadways.
Version 5 of the cartography might have simply listed Main Street as a two lane road with a 30 MPH speed limit, whereas version 8 now shows exactly how many traffic lights or stop signs there are per block, where there are lane restrictions and bottlenecks due to turn lanes, bus bays, and so on. The result is you have more variables available for consideration for the same road, and thus route planning decisions are becoming more granular.
Michael
This oddness over rerouting is not limited to Garmin.
The Blaupunkt system in my car sometimes, when travelling east to west in London, sends me down the Highway to Wapping and home and sometimes via Wapping High Street (which runs parallel to the Highway) and then to home. I have never worked out why.
The Blaupunkt system in my car sometimes, when travelling east to west in London, sends me down the Highway to Wapping and home and sometimes via Wapping High Street (which runs parallel to the Highway) and then to home. I have never worked out why.
Dellis
Registered user
Robin said:
Your perception is correct.
Dave
HMR said:I'm sorry that I have to tell you - No, there is no random element involved!
It was not two identical runs. Fanum did not press the recalculate button on exactly the same spot. Neither did he press the button on exactly the same time of the week. There where probably also GPS signal reception differences. This is enough to explain the different routes selected.
There are other possible reasons. Software bugs for example. A read address bug may make the GPS behave as if a random element was involved. Still, a GPS is a digital computer that operates "exact" and not an analog instrument with an environment dependent precision.
Ok, then. How EXACTLY would he have to repeat the recalculation. To within a second, a millisecond, a microsecond, a nanosecond ? How exactly in the same place would he have to be etc etc. These ARE clearly analogue, continuous variables.
Chaos theory implies that a simple system with only a few input variables can lead to results which are impossible to predict.
In the real world they might be but in order for the computer to make some understanding of them they're sampled into discrete values. It may be that it makes use of ANNs or fuzzy logic but even that's calculated using discrete values.gasman said:
Similar threads
