Data exists
Public transport operations 🚍
General Transit Feed Specification (GTFS)
Data exists
Public transport operations 🚍
General Transit Feed Specification (GTFS)
Data exists
🛣 Infrastructure + 🚍 PT routes
OpenStreetMap (OSM)
Data exists
🚍 Public transport operations
General Transit Feed Specification
(GTFS/GTFS-RT)
🛣 Infrastructure + 🚍 PT routes
OpenStreetMap
(OSM)
Data exists, but…
R package focused on harmonising GTFS and OSM data to enable integrated public transport modeling
Public transport operations 
OSM road network
GTFShift
GTFShift::osm_shapes_match_routes()
Matching of GTFS shapes, \(i\), with OSM relations, \(j\), based on geometrical correspondence using a closeness heuristic
\[C(i, j)\]
that aims to minimize:
GTFShift
GTFShift::osm_shapes_match_routes()
Matching of GTFS shapes, \(i\), with OSM relations, \(j\), based on geometrical correspondence using a closeness heuristic
\[C(i, j) = d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j})\]
that aims to minimize:
- distance between departure and arrival stops, \(d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j})\)
GTFShift
GTFShift::osm_shapes_match_routes()
Matching of GTFS shapes, \(i\), with OSM relations, \(j\), based on geometrical correspondence using a closeness heuristic
\[C(i, j) = d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j}) + |L_{GTFS, i} - L_{OSM, j}|\]
that aims to minimize:
distance between departure and arrival stops, \(d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j})\)
length difference, \(|L_{GTFS, i} - L_{OSM, j}|\)
GTFShift
GTFShift::osm_shapes_match_routes()
Matching of GTFS shapes, \(i\), with OSM relations, \(j\), based on geometrical correspondence using a closeness heuristic
\[C(i, j) = d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j}) + |L_{GTFS, i} - L_{OSM, j}| + \frac{L_{GTFS, i}}{N_{stops, i}} \cdot |N_{stops, i} - N_{stops, j}|\]
that aims to minimize:
distance between departure and arrival stops, \(d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j})\)
length difference, \(|L_{GTFS, i} - L_{OSM, j}|\)
number of stops difference, \(\frac{L_{GTFS, i}}{N_{stops, i}} \cdot |N_{stops, i} - N_{stops, j}|\)
GTFShift
GTFShift::osm_shapes_match_routes()
Matching of GTFS shapes, \(i\), with OSM relations, \(j\), based on geometrical correspondence using a closeness heuristic
\[C(i, j) = d(\text{init}_{GTFS, i}, \text{init}_{OSM, j}) + d(\text{fin}_{GTFS, i}, \text{fin}_{OSM, j}) + |L_{GTFS, i} - L_{OSM, j}| + \frac{L_{GTFS, i}}{N_{stops, i}} \cdot |N_{stops, i} - N_{stops, j}|\]
Shape \(S_i\) is associated with the OSM route \(O_{j^*}\) that minimizes its value:
\[j^* = \operatorname{argmin}_{j} C(i, j)\]
Case study
Lisbon Metropolitan Area
263 GTFS shapes
1.235 trips at peak
60 GTFS shapes
111 trips at peak
130 GTFS shapes
136 trips at peak
1.533 GTFS shapes
2.900 trips at peak
Case study: results
Carris GTFS shape geometries improved after OSM match
Before 
Case study: results
Carris GTFS shape geometries improved after OSM match
Before
After
Case study: results
Going back to the initial example…
Before 
Case study: results
Going back to the initial example…
Before
After
GTFShift
GTFShift::osm_shapes_match_routes()
🌐 Open source method that relies on open data
💯 Simple and effective method to improve GTFS shapes quality
🧐 Enhances accuracy of of trip visualisation
🤓 Relation between PT operations and infrastructure unlocks wide range of modeling possibilities
Motivation
Limit impacts of congestion
(Cesme et al. 2018)
Reducing travel time variability
(Surprenant-Legault and El-Geneidy 2011)
Increase commercial speed
(Basso et al. 2011)
Other modes can loose
quality of service
(Arasan and Vedagiri 2010)
Potential to jeopardize
public acceptance
(Batty et al. 2014)
It is crucial to underlie these interventions on data and evidence to ensure they are effective and accepted by the public
Data sources and criteria
Bus frequency 🚍
General Transit Feed Specification (GTFS)
Traffic conditions 🚗 / Passenger demand 👭
GTFS Real Time (GTFS-RT)
Number of lanes 🛣 / Network continuity 🌐
OpenStreetMap (OSM)
Data sources: GTFS-RT
Traffic conditions 🚗 / Passenger demand 👭
GTFS Real Time (GTFS-RT)
{
"agency_id": "41",
"bikes_allowed": true,
"capacity_seated": 29,
"capacity_standing": 28,
"capacity_total": 57,
"contactless": false,
"id": "41|300",
"license_plate": "AC-45-FH",
"make": "CaetanoBus",
"model": "e.City Gold",
"owner": "SCOTTURB",
"propulsion": "electricity",
"registration_date": "20141110",
"wheelchair_accessible": true,
"bearing": 167,
"block_id": "1_1325-11",
"current_status": "INCOMING_AT",
"event_id": "6KLCD-41|300-1612_0_2_0800_0829_0_1",
"lat": 38.76739501953125,
"line_id": "1612",
"lon": -9.2985200881958,
"pattern_id": "1612_0_2",
"route_id": "1612_0",
"schedule_relationship": "SCHEDULED",
"shift_id": "TP 1+1325",
"speed": 3.611111111111111,
"stop_id": "172486",
"timestamp": 1778574545,
"trip_id": "[6KLCD]1612_0_2_0800_0829_0_1",
"door_status": "CLOSED"
}Methodology
Methodology: Bus frequency 🚍
Aggregate GTFS frequencies at route level and use OSM routes geometry to aggregate at road segment level
GTFShift::get_route_frequency_hourly
GTFShift::osm_shapes_match_routes
GTFShift::get_way_frequency_hourly
Methodology:
Nr of lanes 🛣 / Network continuity 🌐
Getting road network topology from OSM
osmdata::opq
GTFShift::osm_bus_lanes
Methodology: Traffic conditions 🚗
Compute average speed along route geometry for each GTFS-RT update and aggregate per OSM road segment
GTFShift::rt_collect + GTFShift::rt_average_speed
sf::st_buffer
GTFShift::rt_extend_prioritization
Case Study
Lisbon Metropolitan Area
263 GTFS shapes
1.235 trips at peak
1.533 GTFS shapes
2.900 trips at peak
Reproducibility
📦 Open source R package
📊 Comprehensive framework for bus lane prioritization
🔎 Provides methods for the analysis of each individual dimension
🚀 Enables aggregated analysis with one simple method
Reproducibility
🚀 Enables aggregated analysis with one simple method
library(GTFShift)
gtfs = GTFShift::load_feed("https://gateway.carris.pt/gateway/gtfs/api/v2.8/GTFS")
osm_q = opq(bbox=sf::st_bbox(tidytransit::shapes_as_sf(gtfs$shapes))) |>
add_osm_feature(key = "route", value = c("bus", "tram")) |>
add_osm_feature(key = "network", value = "Carris", key_exact = TRUE)
# Bus frequency 🚍, Number of lanes 🛣️, Network continuity 🌐
lanes = GTFShift::prioritize_lanes(gtfs, osm_q)
rt_collection = read.csv("rt_collect_file.csv") |>
sf::st_as_sf(coords = c("vehicle.position.longitude", "vehicle.position.latitude"), crs = 4326)
# Traffic conditions 🚗
lanes = GTFShift::rt_extend_prioritization(
lane_prioritization = lanes,
rt_collection = rt_collection
)
