On 13/02/2008, Stefan Pflumm <[EMAIL PROTECTED]> wrote: > For this i will analyse gps-tracks to calculate the average time the > persons need from one parting node to another dependent of the time of > day. For example: there is a person who needs 30 min from node A to node > B at 17 o'clock. Another person needs for the same distance 20 min at > 17:20. In the routing program i will start to travel at 16 o'clock so > the routing software calculates that i reach the lane A-B at 17:10. > Now the routing software interpolates the time costs from 17:00 and > 17:20 and get a time cost about 25 min.
You'll need to do some pretty smart filtering on the facts in the GPS logs. For instance, if you wish to create statistics for motor vehicles, you'll need to try to exclude pedestrian and bicycle trails. So now you need to see whether you can see enough patterns to tell the difference between the following: * A pedestrian and a car moving very slowly due to congestion * A bicycle and a car moving slowly due to congestion * A voluntary interruption in forward progress (parked, went shopping) and a long time spent standing still for traffic reasons (you shouldn't count the former towards average speed, but the latter you should). Sometimes you'll get clues - a trail that includes some travel at 100km/h isn't a pedestrian or cyclist. One that reaches 20km/h isn't a pedestrian. But sometimes you'll have traces from a hand-held device where the driver or cyclist parked and continued on foot. Not trivial. Dermot _______________________________________________ Routing mailing list [email protected] http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/routing
