[transims-developers] Clarification of Data in Snapshot

[Clark H. <cha...@vo...>]

Hello Sponsors,


We are a team taking on Challenge 4 of the 2020 SMC Data Challenge, and 
we would like clarification regarding possible discrepancies we noticed 
in the traffic simulation snapshot dataset.


Specifically, many of the records in this dataset seem to have position 
data that is inconsistent with both the provided road network and with 
the actual geography of the Chicago area. We’ve identified three main 
problems:

  *

    Over half of the records apparently have link offsets (distances
    travelled) that are greater than the actual lengths of their
    associated links.

  *

    On top of that, many records have UTM coordinates that lie inside of
    buildings, on train tracks, in rivers, or are even in Lake Michigan.

  *

    Even when the link offset within a record is valid, it often does
    not match the UTM position in the record: computing the absolute UTM
    position of the vehicle from the link and offset information results
    in positions that can be hundreds or thousands of meters away from
    the stored UTM position.


One specific example of the first problem can be seen in row 55 of file 
Snapshot_2000000.csv, describing vehicle 34970 at time 7:24. It is 
listed as travelling down link 1210 with an offset of 202.5 meters; 
however, link 1210 should only be 14.6 meters long, based on the 
coordinates provided in the road network dataset.

After taking a random sample of roughly 5% of the dataset (1.2 million 
records), we found that almost 685 thousand records (or about 55% of the 
sample) have this problem.


An example of the second and third problems can be seen in row 105,310 
of file Snapshot_4000000.csv, which describes vehicle 34271 at time 
9:00. This record contains a valid link and an offset that is consistent 
with the computed length of that link; however, the given UTM 
coordinates for this vehicle (16N 447301 4635366) place it in the middle 
of the Chicago River.

We tried computing a new position for the vehicle based on the link and 
offset information provided in the record, by using the offset to 
interpolate between the endpoints of the associated link. However, the 
position as calculated by this method is 1086 meters away from the 
position stored in the record.


Furthermore, we’d like to ask if parked cars are simulated by TRANSIMS, 
and if that might cause simulated vehicles to appear to be inside of 
buildings.


Thank you for your time.


Sincerely,

Sebastian Mobo and Clark Hathaway