agentPlans

C++17 port of the two TSM "trip-list" R scripts from the QGIS tsm_panel plugin:

R script (canonical source)	agentPlans stage
Rscripts/2_Create_LDT_TripTable_GA_AL_template.R	ldt
Rscripts/3_get_ELTOD_TripTable_template.R	eltod

Canonical R source: C:/Users/<user>/AppData/Roaming/QGIS/QGIS3/profiles/default/python/plugins/tsm_panel/Rscripts/

The two scripts build the long-distance (LDT) trip list and then combine it with the SDT resident/visitor trip lists and the truck ODME table into the ELToD trip list that Hydra (AgentFlow DTA) consumes. agentPlans runs the full pipeline from a single control file and writes the trip list as a gzipped CSV in the exact Hydra schema.

Build

Needs MSVC (VS18 / VS 2026) + Ninja + CMake. No external dependencies — gzip I/O uses vendored miniz (the same library Hydra uses).

scripts\build.bat

Produces build\agentPlans.exe.

Run

build\agentPlans.exe config\agentplans_settings.txt

Inputs

All inputs are plain CSV. The original Excel lookups must be exported to CSV once — see docs/CSV_INPUTS.md for the exact schema of every file (column names are matched by header). Paths are set in the control file (config/agentplans_settings.txt).

Outputs (written under scenario_dir)

file	meaning
ELTOD_tt_List_hourly.csv.gz	Hydra trip list — primary deliverable
ELTOD_tt_HourClock.csv	combined ELToD OD trip table (wide, by market)
ELTOD_SDT_Res_hourly.csv	SDT resident purpose×VOT OD table

The trip-list gz uses the 14-column Hydra schema, in order:

hh_id,person_id,tour_id,trip_id,valueOfTime,purpose,depart_time,O,D,marketVot,vehTrips,occupancy,hhIncome,market

depart_time is an HH:MM:SS clock (start of the period bin) — Hydra's tsm_trip_reader reads this as its preferred clock departure field. Bin width follows output_resolution (30-min → :00/:30, 15-min → :00/:15/:30/:45).

Fidelity notes (read before comparing to R output)

• RNG: R's sample()/set.seed() stream cannot be reproduced bit-for-bit in C++. Stochastic assignments (external stations, airport zones/modes, time-of-day, truck periods, truck VOT) are statistically equivalent and reproducible for a fixed seed, but individual rows differ. Aggregate trip tables match within sampling noise.

• period column / temporal resolution: every source is brought onto a common internal 15-min grid (LDT and trucks are daily and get timed via the 15-min ToD share table; SDT 30-min periods are split to 15-min), then the output period is collapsed to output_resolution. The bins are declared in the control file:

  setting	meaning	values
  output_resolution	trip-list depart_time bin width	15 or 30 (minutes)
  sdt_input_resolution	SDT trip-list native bin	15 or 30
  ldt_input_resolution	LDT native bin	daily
  truck_input_resolution	truck ODME native bin	daily

  The trip list's depart_time is an HH:MM:SS clock at the start of the output bin, so Hydra's tsm_trip_reader reads it directly (its preferred clock field). The R script wrote a 1..96 15-minute integer index instead; agentPlans makes the resolution explicit and emits a real clock. Hourly OD tables always use the 15-min segment → clock-hour mapping as R.

• LDT/truck rows: the R script collapses LDT tours with .N (count) into vehTrips; agentPlans emits one list row per tour with vehTrips = 1 (identical totals, more rows — natural for a trip list). Truck rows are one per O-D-class with vehTrips = the (scaled) matrix value.

• Not ported: LimitCounties (needs hhID_by_county.csv), telework (needs person_Industry_type.csv), and the TRANSIMS export are guarded off; LimitCounties=true raises an error. Pure diagnostic dcast/print summaries from the R scripts are omitted.

Layout

src/util/   csv, gzip read/write (miniz), RNG + qnorm
src/        settings (control file), lookups, stage_ldt, stage_eltod, main
config/     control-file template
docs/       CSV input schemas
third_party/miniz   vendored gzip codec