Coding error in Navigations and some suggestions

Bugreport for Ovi Maps 3.0 Beta for mobile

21-06-2009 22:36

Today I tried out the navigation properly and here are some of what I noticed: -

1 - Once you ask to be guided to a destination a route is planned, and if you deviate from that route, you are guided back to the route instread of a total route recalculation. This is poor because there may be a good reason to avoid a route. In my example we were going half way to our destination and nokia maps was telling us to go BACK to where we came from to get back onto its route. pathetic. Whoever coded this, this IS NOT smart coding, code it so that it recalculates the route in total. When I stopped navigation and got it to calculate the route again, it stopped asking me to go back.

2 - In order to make up for not dealing too well with gps data and to always be on the same route as is planned, nokia maps obviously has some tolerance function that if you are within some distance x from the road, you are assumed to be on the road, and perhaps some other little tricks. A problem with this is that when you deviate from the route, it takes several seconds before the software 'realises' you are on a different route. Improve how you deal with gps data and reduce the tolerances!

3 - The actual navigation, that is the routing, is poor, it so happens that we know where we are going, yet sub optimal, or just plain bad routes are suggested by nokia maps. I cant help you fix your routing without working on this project with you! So get some smarty pants to fix it!

Thank you and good night.

PcBorg

Resistance... Is mandatory.

Comments

Submitted by navio1973 on 21-06-2009 23:29.

I agree with you about the 3rd point. OM3 doe not calculate always the right route as I wrote several times. If you use viamichelin, googlemaps, livemaps or other programs they show route complety different, but right, from OM (website) and OM3
Navio
Nokia E66-1 (218.04) FW 200.21.118

Submitted by Fonero on 22-06-2009 15:54.

pcborg wrote:1 - Once you ask to be guided to a destination a route is planned, and if you deviate from that route, you are guided back to the route instread of a total route recalculation.

This is not my experience when OM3 has recalculated the route for me. Of course in some cases the best route is to come back to the originally suggested one, but in most cases OM3 has not suggested that to me.

pcborg wrote: 2- ...if you are within some distance x from the road, you are assumed to be on the road

Yes, that is the case. However I think that is unavoidable. The accuracy of GPS simply isn't that good. Accuracy is within a few meters if you are boating with full view to the sky. In a car the signal is blocked from most directions by the chassis of the car, by nearby buildings etc. If the gps used the raw coordinates then it would be constanly telling you to "return to the nearest road".

Markus
(E66-1(03), 210.21.007, Maps 3, v3.01 09wk24 b01 MW OpenGL LM MN, Map version 0.1.22.103)

Submitted by oolaoden97 on 22-06-2009 18:09.

@Fonero, I agree with pcborg. This happens to me too. In fact, it happened today again and I had previously opened a thread about this. In my handset, Nokia maps actually highlights both the original route and the recalculated route. So, while the voice nav tells me to turn right, the screen is also showing me I can go straight on. Very confusing! Let's not treat pcborg's concerns (and mine) as some kind of an isolated incident so that it will get the proper attention it deserves by the Nokia maps guys and they can fix it. I truly believe this is a serious bug.

Submitted by ErikEngerd on 22-06-2009 20:35.

I agree too that points 1 and 3 are critical as these are the core functionality of car navigation.

For 1 there are probably many options already in the literature. One solution I can think of is to artificially manipulate the map data in memory to lower the velocity that is used in route calculation for roads that you already traveled on. That is a simple trick that will avoid this phenomenon in many cases by making it unattractive to do U turns and return to the same route. As a slightly more enhanced version, the velocity should also be artificially lowered on the next km or so of the original planned route that was missed because of the deviation.

For number 3 I guess the Nokia maps team would like to have concrete examples so they can verify the problem.

As far as point 2 is concerned, I think that is a tuning issue. Basically, it is a good thing that maps takes gps inaccuracy into account, but there must be a balance between dealing with this inaccuracy and being in time for recalculation. In TomTom it also takes a few seconds before it realises you're off track, and occusionally it is too slow in presenting a modified route but in 99% of cases it is on time.

As one extreme case I can mention the first mapsonic versions on Palm OS that suffered from various problems
ad 1. It would frantically avoid doing U turns or using the same roads again. Probably it is an extreme case of reducing the velocity on already visited roads to near 0. That resulted in large detours in cases where a U turn or short drive around the block to return to the same route would have been better.
ad 2. It did not have any mechanism to deal with inaccuracy of the GPS or map material (!) resulting in interesting voice commands like "turn sharp right" on the highway (scary!).

Another case was digimap (also Palm OS) that was too slow in detecting being off route and also way too slow in recomputation of the route, leading to a lot of despair driving in a big city and making a mistake.

Submitted by ErikEngerd on 22-06-2009 20:55.

Recalculation of routes can also be done by using the point where the user deviated from the route and then adding a fixed distance to it along the previous route to determine a via point. If the remainder of the route is shorter than this fixed distance, then take the end point of the route as via point. Then instead of full rerouting, route to that via point and from there connect it to the remainder of the previous route.

In that way, full recalculation logic is there, but the effort in doing recalculation is limited. It is also a nice tuning parameter. For slower devices, the fixed distance should be chosen small and faster devices can use a larger value.

The advantage of this approach is that you don't need a separate algorithm for recalculation but that you can use the original high-quality algorithm with reduced overhead because of the above limit of the fixed distance.

Submitted by Fonero on 23-06-2009 12:50.

There are some instances where routing seems in my eyes suboptimal.

1 Near my summer cottage I have a situation where I come to a crossing, drive straight 150 metres (speed limit 50 km/h) and arrive at my destination (local shop). OM3, however, advices me to turn left on a bigger road (80 km/h) and drive 150 m, then turn right and drive 50 m along a smaller road (50 km/h) and then arrive to destination.

This is right advice counting distances and speed limits, but in practise turning, accelerating and breaking takes extra time.

2 I drive along a major road (2 lanes, not motorway). The road slowly turns right. Then OM3 advices me to take a small road on the right and drive a few hundred meters along it. Both roads have the speed limit 80 km/h and the small one is a few meters shorter.

I suppose OM3 doesn't recognize the difference between major and minor roads (if not motorways) if their speed limits are the same.

What could be done? I don't know what information is in map data. In case 2 there should be info on a fine-grained classification of roads.

In case 1 OM3 should add extra seconds for accelerating and breaking at crossings and sharp turns. Adding penalty for crossings might help in case 2 too.

Somebody mentioned that OM3 suggest driving through Helsinki instead of using the ring-road. Adding a penalty for sharp turns and crossings might help there too. Better solution might be to attach two speeds to city streets: official speed limit of 50 km/h (or 40 km/h) and a slower pracical driving speed.

Markus
(E66-1(03), 210.21.007, Maps 3, v3.01 09wk24 b01 MW OpenGL LM MN, Map version 0.1.22.103)

Submitted by ish68 on 28-06-2009 16:29.

I have our local tesco extra as a favourite, so when we're out of town it takes us there (baby likes hearing the voice LOL), but strange things happended yesterday.

I set the destination as tesco and the destination it waned me to end at was about 2 miles from tesco. The location mark on the phone was on the road at the side of tesco for the 2 miles and it kept telling me to take a U-turn even where U-turn's are not allowed. This destination has worked all the way throughout maps 2 and the earlier beta of maps 3.

It was as though the GPS knew where I needed to be and told me I was there and then as I got nearer to the location OMM3 was saying telling me I was not... Errrr!!

When at tesco I reset my location to tesco and chose 'drive to' and it set I was 105 yards from the location, so I know it was the correct co-ordinates received from my favourites.

Is this yet another navigation problem?

Ish68
UK Vodafone - N95 8Gb - Code: 0557135 - FW(31.0.015)