The breakdown = TRUE output of expanded_travel_time_matrix() gives data on access_time, wait_time, ride_time, transfer_time, egress_time, routes, n_rides and total_time. Some of these metrics represent collapsed per-leg information into summed metrics (e.g total wait time and total ride time). This is sufficient for some applications, but loses valuable information in the context of activity-based travel demand modelling, for which we are currently trialing r5r.
Specifically, I would like to propose exposing:
- Per-leg wait times. The time spent waiting at each "transit stop" before boarding. Alternatively a derived split into initial wait (waiting at the transit stop before boarding the first bus/rail/tram etc.) vs. transfer wait (all subsequent wait times) would also be valuable information
- Per-leg in-vehicle (ride) times: The time spent on each individual route. Alternatively a breakdown by "mode". So if I define BUS and RAIL as valid routing alternatives, then for each routing I would like to know the ride time on the/all busses and the ride time on the/all trains.
Happy to look further into how to implement this. In our specific case this would make r5r a very good alternative to commercial software to create the skims for our travel behaviour model (i.e. ActivitySim). Any sophisticated mode choice utility specification will however require a more detailed breakdown of the time related impedance metric between OD zones. We have also trialed the detailed_intiraries() function, but have found it to be prohibitively computationally expensive for large OD matrix calculation.
One extra note:
I think it would be valuable if you would more concretely document what the metrics represent. Especially "transfer_time" can be quite ambiguous if this refers to time spent walking (as I understand it does) or waiting (for the next bus to arrive) or the sum of both.
Thanks for maintaining r5r!
The breakdown = TRUE output of expanded_travel_time_matrix() gives data on access_time, wait_time, ride_time, transfer_time, egress_time, routes, n_rides and total_time. Some of these metrics represent collapsed per-leg information into summed metrics (e.g total wait time and total ride time). This is sufficient for some applications, but loses valuable information in the context of activity-based travel demand modelling, for which we are currently trialing r5r.
Specifically, I would like to propose exposing:
Happy to look further into how to implement this. In our specific case this would make r5r a very good alternative to commercial software to create the skims for our travel behaviour model (i.e. ActivitySim). Any sophisticated mode choice utility specification will however require a more detailed breakdown of the time related impedance metric between OD zones. We have also trialed the detailed_intiraries() function, but have found it to be prohibitively computationally expensive for large OD matrix calculation.
One extra note:
I think it would be valuable if you would more concretely document what the metrics represent. Especially "transfer_time" can be quite ambiguous if this refers to time spent walking (as I understand it does) or waiting (for the next bus to arrive) or the sum of both.
Thanks for maintaining r5r!