Incorrect activity type suggestion

[tvage]

When I use GPS Logger for recording cycling it always suggests the activity is driving despite the average speed being somewhere around 16 miles per hour. I suspect that the maximum speed detected is being used to determine the activity and that as this could easily be in excess of 40 mph for short downhill sections GPS Logger thinks this is a driving speed. Can you please review the way that activities are suggested?

Thanks,
Mike

[GrazianoCapelli]

Thank you for opening this issue and sharing your experience.
It is very difficult to accurately distinguish whether a track was recorded by bike or car, because we can record tracks with the same characteristics using both means of transport.
Currently, the app calculates the average of the average moving speed and the maximum speed (speedAverageMoving + speedMax) / 2 and decides the suggested activity type based on this value.
It is not 100% exact, but it returns in general a good result in our tests.

You can find the calculation here:

GPSLogger/app/src/main/java/eu/basicairdata/graziano/gpslogger/Track.java

Line 946 in 963f1d6

public int getEstimatedTrackType() {

We are open to review this evaluation method; any suggestions or ideas for improving the discrimination, also using other tracks data, are more than welcome.

[tvage]

Hi thanks for the link to the code used to determine track type. I ride socially once a week with a club in a group of mainly pensioners and do no training so am by no means fast, but looking at the statistics for my rides almost every ride has a max speed in excess of 50 km/h, with a few in excess of 60 km/h and occasionally exceeding 70 km/hr. The average moving speed of most of my rides is over 20km/hr but under 25km/hr. More committed club riders will likely be in the 25-35km/h range.

Changing:
if (speedMax < (50.0f / 3.6f)) {
if ((speedAverageMoving + speedMax) / 2 > 35.0f / 3.6f) return TRACK_TYPE_CAR;
to:
if (speedMax < (70.0f / 3.6f)) {
if ((speedAverageMoving + speedMax) / 2 > 40.0f / 3.6f) return TRACK_TYPE_CAR;

would correctly evaluate most of my rides (but still give incorrect results for faster riders). However, I'm unclear why the code uses the midpoint of the max speed and average moving speed in the calculation. In my view this makes it more difficult to determine the track type. A bicycle with an average speed of 20km/h and a maximum speed of 50km/h would have a midpoint of 35km/h as would a car ar a pretty much constant 35km/h in a 20mph built up area zone, as in towns in Wales. Using this figure does not allow us to recognise that although both scenarios could be either a bicycle or a car, the first scenario is more likely to be a bicycle and the latter a car.

A much better idea of track type would be obtained by comparing the speed profile with the gradient profile. If the speed slows on uphill and increases on downhill it is almost certainly a bicycle. If there are lots of stops then it is probably a car in a traffic jam.

If you don't have the data to evaluate the number of stops, elevation changes and the duration of the activity at this point in the code then I think the commented out code which just uses the average moving speed would give better reults (with the limit set to 25 instead of 20) than the current midpoint code. You could perhaps add a user setting for the maximum average speed for a bicycle which would allow a fast club cyclist to set a high limit and a casual rider on a city bike to set a low one.

Regards,
Mike