Frequency-based trips

[irees]

Handle frequency.txt style trips, e.g.
https://rawgit.com/CUTR-at-USF/bullrunner-gtfs-realtime-generator/master/bullrunner-gtfs.zip

[irees]

Investigate multiplying-trips vs. internally supporting frequency.txt based trips.

[barbeau]

I would suggest taking a look at OpenTripPlanner, as they've dealt with this same question. I haven't delved into the code myself, but a student from our group did some prototyping in OpenTripPlanner for supporting frequency-based exact_times=0 real-time info (opentripplanner/OpenTripPlanner#1647), and also created a frequency-based exact_times=0 GTFS-realtime feed for the USF Bull Runner (https://github.com/CUTR-at-USF/bullrunner-gtfs-realtime-generator).

[drewda]

@irees @gknisely while this ticket waits on the backburner, likely until next quarter, let's downgrade all frequency-based feeds to import level 2, so they are not imported into Valhalla for routing.

[irees]

13 currently active feeds that use frequencies.txt -- trip_count is the total number of trips.txt trips, frequencies_count is the number of frequencies.txt entries.

[{
    "onestop_id": "f-drt-mbta",
    "active_feed_version": "9c1b5facfed0b7cab2742492124f42989d487b7b",
    "operators_in_feed": [{
        "onestop_id": "o-drt-mbta",
        "name": "MBTA"
    }, {
        "onestop_id": "o-drt3p-massport",
        "name": "Massport"
    }],
    "trip_count": 158417,
    "frequencies_count": 21
}, {
    "onestop_id": "f-f25d-socitdetransportdemontral",
    "active_feed_version": "b3cb8fd3828ef9dcabc07b4bd5b86cdbd3478f02",
    "operators_in_feed": [{
        "onestop_id": "o-f25d-socitdetransportdemontral",
        "name": "Société de transport de Montréal"
    }],
    "trip_count": 127566,
    "frequencies_count": 1032
}, {
    "onestop_id": "f-9qh0-anaheim~ca~us",
    "active_feed_version": "c93d84a5208adc5d4a4e6729a10a5d35c24683d0",
    "operators_in_feed": [{
        "onestop_id": "o-9qh0-anaheimresorttransportation",
        "name": "Anaheim Resort Transportation"
    }],
    "trip_count": 13616,
    "frequencies_count": 1
}, {
    "onestop_id": "f-s-anm~it",
    "active_feed_version": "0977871b4b1767955025d9e90853c4326b71146a",
    "operators_in_feed": [{
        "onestop_id": "o-s-aziendanapoletanamobilit",
        "name": "Azienda Napoletana Mobilità"
    }],
    "trip_count": 4756,
    "frequencies_count": 1573
}, {
    "onestop_id": "f-9g3-semovi",
    "active_feed_version": "fd847780ff1c19fdc45756fa47a489a5c0d75b01",
    "operators_in_feed": [{
        "onestop_id": "o-9g3r-ferrocarrilessuburbanos",
        "name": "Ferrocarriles Suburbanos"
    }, {
        "onestop_id": "o-9g3q-serviciodetransporteselctricos",
        "name": "Servicio de Transportes Eléctricos"
    }, {
        "onestop_id": "o-9g3-reddetransportedepasajeros",
        "name": "Red de Transporte de Pasajeros"
    }, {
        "onestop_id": "o-9g3q-nochebscorredoresconcesionados",
        "name": "NOCHEBÚS Corredores concesionados"
    }, {
        "onestop_id": "o-9g3w-sistemadetransportecolectivometro",
        "name": "Sistema de Transporte Colectivo Metro"
    }, {
        "onestop_id": "o-9g3w-metrobs",
        "name": "Metrobús"
    }, {
        "onestop_id": "o-9g3q-corredoresconcesionados",
        "name": "Corredores Concesionados"
    }],
    "trip_count": 1679,
    "frequencies_count": 1679
}, {
    "onestop_id": "f-kzf-nairobi~saccos",
    "active_feed_version": "fdb6aad5e0c21c010414eed2e7245786330911ad",
    "operators_in_feed": [{
        "onestop_id": "o-kzf-nairobi~saccos",
        "name": "Approved SACCOs of Nairobi as mapped by University of Nairobi C4D Lab"
    }],
    "trip_count": 268,
    "frequencies_count": 804
}, {
    "onestop_id": "f-9w28-naipta~az~us",
    "active_feed_version": "0e475297f9e8842754f447f74ec256fa9fa68520",
    "operators_in_feed": [{
        "onestop_id": "o-9w28-mountainline",
        "name": "Mountain Line"
    }],
    "trip_count": 125,
    "frequencies_count": 126
}, {
    "onestop_id": "f-dnjj-clemson~sc~us",
    "active_feed_version": "db1eac59d735f378629c47a18722f0bff56a7425",
    "operators_in_feed": [{
        "onestop_id": "o-dnjj-clemsonareatransit",
        "name": "Clemson Area Transit"
    }],
    "trip_count": 94,
    "frequencies_count": 45
}, {
    "onestop_id": "f-69y7-recursosdatabuenosairesgobar",
    "active_feed_version": "95742a1a65826abfa23f5aeaeeb082e45f066801",
    "operators_in_feed": [{
        "onestop_id": "o-69y7-sbase",
        "name": "Subterráneos de Buenos Aires"
    }],
    "trip_count": 48,
    "frequencies_count": 172
}, {
    "onestop_id": "f-9xhv-nps~romo~shuttles",
    "active_feed_version": "55b23bcff5ccf13a08e1b1fbe49308e34f086ebd",
    "operators_in_feed": [{
        "onestop_id": "o-9xhv-rockymountainnationalpark",
        "name": "Rocky Mountain National Park"
    }],
    "trip_count": 16,
    "frequencies_count": 16
}, {
    "onestop_id": "f-u3qc-metrowarszawskie",
    "active_feed_version": "3a7d41d38edf4e1d1d3340103cc6ab5245e3128b",
    "operators_in_feed": [{
        "onestop_id": "o-u3qc-metrowarszawskie",
        "name": "Metro Warszawskie"
    }],
    "trip_count": 16,
    "frequencies_count": 98
}, {
    "onestop_id": "f-u0-openov~lu",
    "active_feed_version": "72c96e0ef267db148ab09ea6d928035e258f9435",
    "operators_in_feed": [{
        "onestop_id": "o-u0u1-tice",
        "name": "TICE"
    }, {
        "onestop_id": "o-u0u-rgtr",
        "name": "Régime Général des Transports Routiers"
    }, {
        "onestop_id": "o-u0-cfl",
        "name": "Société Nationale des Chemins de Fer Luxembourgeois"
    }, {
        "onestop_id": "o-u0u6-avl",
        "name": "Autobus de la Ville de Luxembourg"
    }],
    "trip_count": 13789,
    "frequencies_count": 947
}, {
    "onestop_id": "f-dhvrs-bullrunner",
    "active_feed_version": "3e3efb8f611d9189ebe819aec85ba030ab6ae8b4",
    "operators_in_feed": [{
        "onestop_id": "o-dhvrs-usfbullrunner",
        "name": "USF Bull Runner"
    }],
    "trip_count": 15,
    "frequencies_count": 15
}]

[barbeau]

Are you all currently supporting frequency-based exact_times=1 trips (i.e., the compressed representation of schedule-based trips)? Or are these on back burner for now as well?

[irees]

My current concept is to describe frequency-based trips using "ScheduleStopPair" attributes "frequency_headway_seconds", "frequency_start_time", and "frequency_end_time", with arrival & departure times relative to the start of the trip. I don't currently have a flag for exact_times but it could be added.

For example, an SSP with frequency_headway_seconds=600, frequency_start_time=08:00:00, and origin_departure_time=00:05:00 would result in departures at 08:05, 08:15, 08:25, etc.

We've been discussing this lately, and I welcome any feedback and suggestions.

[barbeau]

@irees I'm not very familiar with the internal data model of transitland, but what you describe seems to make sense for feeds with exact_times=1. This is effectively a compressed way to represent a fixed schedule without including a bunch of near duplicate trips, and most transit software I've seen treats it exactly as you describe.

However, exact_times=0 (or if exact_times is omitted) is a different beast. This is true frequency-based, meaning that the expected arrival time of the vehicle shifts throughout the day, based on when the vehicle actually arrives.

This has a lot of implications when you start getting into user-facing information, such as answering the question "when is my bus expected to arrive?" For exact_times=1, the answer is the same as normal scheduled trips. For example, assuming lack of real-time info, if it's 8:20, "your bus is scheduled to arrive in five minutes". But for exact_times=0, the answer is "your bus is expected to arrive every 600 seconds." And, if you don't have real-time info, that answer is the same no matter when during the day you ask the question. For routing and calculating wait times, without real-time info systems like OTP use a pessimistic estimate and assume the wait is always the headway. If you have real-time info with a predicted arrival time, then you materialize an estimated arrival time on the fly based on this estimate - in that case, then you can say "your bus is expected to arrive in five minutes".

This seems to be a subtle difference, but in creating a GTFS-rt frequency-based exact_times=0 feed, and trying to add support for this into OTP, we ran into a lot of unexpected issues. IMHO, true frequency-based systems are tricky.

[drewda]

@irees the one final step that can happen on another issue: adding mention of the new frequency attributes to the SSP docs: https://github.com/transitland/www-transit-land/issues/274

[irees]

Frequency-based schedules are now supported in the datastore & in routing.

Example:
https://valhalla.mapzen.com/route?json={%22locations%22:[{%22lat%22:28.05732490546724,%22lon%22:-82.40939140319824,%22type%22:%22break%22,%22name%22:%22%22},{%22lat%22:28.080386371114074,%22lon%22:-82.41333961486816,%22type%22:%22break%22,%22name%22:%22%22}],%22costing%22:%22multimodal%22,%22costing_options%22:{%22transit%22:{%22use_bus%22:%221.0%22,%22use_rail%22:%220.0%22,%22use_transfers%22:%221.0%22}},%22directions_options%22:{%22language%22:%22en-US%22},%22date_time%22:{%22type%22:0}}
(add your api key)

[barbeau]

@irees Nice! So what assumption does Valhalla make for the estimated wait time for the bus for a frequency-based exact_times = 0 system like our USF Bull Runner? IIRC OTP uses a pessimistic approach and assumes you'd need to wait the entire headway defined in GTFS.