Handle optimistic routing engine travel times for cars

[Hussein-Mahfouz]

Travel time matrices from r5 use road speed limits for car routing (r5 does not account for congestion). Our travel time matrix overestimates how far you can go in a given time. Activities may therefore be assigned to locations that are further away than they should be. We should:

• Validate: The NTS has reported time TripTotalTime and distance TripDisIncSW. Plot TripTotalTime vs TripDisIncSW for NTS data, and compare it to TripTotalTime vs ActualDistance in the matched data (ActualDistance is calculated after assigning an activity to a location). How do the results compare? Any differences by mode?
• Suggestions for fixes:
  • Edit osm pbf input to routing engine by addign congestion. Example of editing pbf file here
  • Compare our results to google travel time matrix api - can we scale our travel times up by a factor?
  • Use distance instead of time? "Previous research [47] found that Minkowski distance with a coefficient of 1.54 best approximates road distance. To identify potential locations for A, we thus searched for locations of the desired type situated in a Minkowski distance (coefficient 1.54) between min trip distance and max trip distance from A-1"