Individual Trip and Vehicle timestamps

[Evan Siroky]

Hello All,

In discussions with MTC staff, we at Caltrans discussed what we believe to be a problem with how timestamps are generated for individual vehicles and trips. MTC staff has confirmed that the timestamps for individual trips and vehicles are not passed through from precursor GTFS-RT systems and instead are generated at the time which MTC receives an entire precursor feed. We believe this design decision is in conflict with the GTFS-RT best practice about Vehicle Positions which states:

Providing the timestamp at which vehicle position was measured is strongly recommended.

At Caltrans, we download data from precursor feeds as well as all MTC subfeeds. We have observed that the individual timestamps coming from the MTC system vary less dramatically than they do in precursor feeds and are also consistently lower. For a number of transit agencies, the difference between the individual timestamp at the precursor feed and the MTC subfeed is sometimes numerous orders of magnitude different. I am attaching a few charts that demonstrate what we are seeing.

We wanted to bring this to the attention of the larger community to see whether this is a concern to anyone. At Caltrans, we sometimes only have access to MTC's feed for a few transit agencies so we rely exclusively on this data in a few instances. Therefore, we are hoping that it can be as accurate as possible.

Thanks,

Evan Siroky

Research Data Manager

Transit Data Quality at Caltrans

[Juan Borreguero]

Hi everybody,

This is indeed a concern for us (Transit App) since we are displaying inaccurate information to the riders.

We consume the 511 regional feed in order to display real time vehicle positions and trip updates: 

https://api.511.org/Transit/VehiclePositions?agency=RG

https://api.511.org/Transit/TripUpdates?agency=RG

These erroneous timestamps are annoying for the riders since they will not have accurate information of where the vehicles are.

In the app, we display the age of the vehicle location (as now - timestamp) and, given this error in the feed, the users will see the wrong age.

This particular example is for Petaluma transit which, according to the analysis above, has a possible error of 2.5 min. In this case, the rider will think the vehicles were there ~30 seconds ago while in reality they were there 3 minutes ago.

It is true that the predictions should give them the information of when the bus is coming to their stop but, according to the feedback we receive, we know for certain that many people rely more on the vehicle locations than on the predictions to plan their trips. 

Related to this point, we ignore if the erroneous timestamps have an impact on the quality of the predictions that the 511 feed is producing.

As a final note to confirm the behaviour expressed in the message above, we constantly measure different stats of the vehicle positions feeds that we consume and, if we take the vehicles of a particular agency from the 511 feed, we notice that they all update at unison with the same timestamp.

We would like to add our voice to the concern above. It would increase the quality of the data and ultimately be beneficial for riders if this issue was fixed.

Thanks for bringing this matter to this group and to the community.

Best

Juan Borreguero

Real Time Data Lead at Transit