gps2gtfs

Website: https://e-kotov.github.io/gps2gtfs/

gps2gtfs turns raw GPS pings from public transit vehicles into GTFS-shaped trips and stop_times tables. It cleans the pings, cuts each vehicle's trajectory into trips between terminals (or at layovers, or by a trip annotation that the pings already carry), detects stop visits with arrival, departure and dwell times, and traces the geometry each trip actually drove. It is a high-performance R port of the Python gps2gtfs package.

Where it fits

GTFS-RT Vehicle Positions ──(gtfsrealtime)──┐
                                            ├─► gps2gtfs ─► trips, stop_times, shapes
raw AVL / logger GPS ───────────────────────┘                     │
                                                                  ▼
                                   gtfsrt2static ─► realized GTFS zip ─► gtfsio / gtfstools / tidytransit

gtfsrealtime parses GTFS-Realtime protobuf. Its Vehicle Positions output uses the column names gps2gtfs expects (vehicle_id, latitude, longitude, timestamp, speed), so it feeds the pipeline with no adapter.
This package does the spatial inference: trips, stop visits and driven shapes from coordinates.
gtfsrt2static assembles the result into a standard-compliant static GTFS feed, either inheriting the identifiers of a planned feed or scaffolding one from scratch.

Installation

# install.packages("pak")
# pak::pak("e-kotov/gps2gtfs")
install.packages("gps2gtfs", repos = c('https://e-kotov.r-universe.dev', 'https://cloud.r-project.org'))

A Rust toolchain is optional. When cargo and rustc are found at install time, the heavy steps are compiled in Rust; otherwise a C++ (Rcpp) backend is built. backend = "auto" prefers Rust, then Rcpp, then a pure R implementation that uses data.table and sf. All three return identical tables.

Example

The package ships one day of GPS pings from a bus on route 654 between Kandy and Digana, Sri Lanka, with the route's stops and its two terminals:

library(gps2gtfs)

result <- g2g_extract_trips_and_stop_times(
  gps_data = g2g_data_gps,
  terminals_data = g2g_data_terminals,
  stops_data = g2g_data_stops,
  terminals_buffer_radius = 50,
  stops_buffer_radius = 30,
  stops_extended_buffer_radius = 50,
  # each stop-direction label maps to the terminal its trips start from
  stop_direction_map = c("Kandy-Digana" = "BT01", "Digana-Kandy" = "BT02"),
  return_trajectory = TRUE
)

result$trips          # one row per trip: terminals, direction, start/end time
result$stop_times     # one row per stop visit: arrival, departure, dwell
g2g_diagnostics(result)                       # what was kept and dropped, by stage
shapes <- g2g_shapes_from_trips(result$trajectory)  # shapes.txt-shaped table

All times are absolute POSIXct values, never clock strings, so a trip that runs past midnight stays unambiguous.

Learn more

Get started: the whole pipeline on the bundled data, step by step.
GTFS-Realtime Vehicle Positions to GTFS: reading archived feeds with gtfsrealtime, deriving stops and terminals from a planned feed, and the trip-annotation fast path.
From raw GPS to a realized GTFS feed: the end-to-end walkthrough across gtfsrealtime, gps2gtfs, gtfsrt2static and gtfstools, on the gtfsrt2static website.
Function reference.

Credits