Configuration¶
Each engine reads one properties file. The keys below are the ones that switch on MILES's models; the values are
those of the final runs. The SILO settings of the final runs are collected in landuse/run/adopted_settings.sh, and
miles landuse … sets the model switches for you.
Land use (SILO)¶
Written by landuse/run/write_properties.py <scenario> <start> <end> [skim_years] [seed] [spec].
| Key | Value |
|---|---|
base.year, start.year, end.year |
2016, 2016, 2024 |
random.seed |
1 to 5 |
number.of.threads |
12 |
travel.time |
SKIM (fixed skims) |
auto.peak.sov.skim.2016 |
auto_peak_time.csv.gz |
transit.peak.time.2016 |
transit_peak_time.csv.gz |
household.file.ascii … job.file.ascii |
the 2016 base hh, pp, dd, jj |
| Key | Value |
|---|---|
resloc.twolevel |
true (PUMA, then dwelling) |
resloc.twolevel.puma.coefficients |
input/residential_puma_coefficients.csv |
resloc.twolevel.dwelling.coefficients |
input/residential_dwelling_coefficients.csv |
resloc.twolevel.puma.zone.file, resloc.twolevel.mig.zone.file |
zones of each PUMA and migration PUMA |
resloc.twolevel.zone.attributes.file, resloc.twolevel.zone.core.file |
zone attributes; DC and Baltimore City zones |
resloc.twolevel.puma.adjacency.file |
PUMA adjacency |
resloc.marketsize, resloc.turnover.file |
stock_turnover; turnover rate by PUMA |
resloc.puma.vacancy.availability |
true (PUMA weight scaled by its vacancy rate relative to 2016) |
resloc.race.share.fixed2016 |
false (own-race shares recomputed every year) |
population.control.until.no.vacancy |
true |
resloc.move.duration |
−0.982 |
resloc.iv.dwelling.sample |
200 |
| Key | Value |
|---|---|
car.ownership.coefficients.file |
input/car_ownership_estimated_coefficients.csv |
car.ownership.region.constants.file |
input/car_ownership_county_constants.csv (the estimated Baltimore City term; 0 elsewhere) |
| Key | Value |
|---|---|
job.choice.coefficients.file |
input/workplace_choice_coefficients.csv (24 terms; District term as estimated) |
job.choice.size |
jobs (all job types weighted equally) |
job.change.rate |
0.252 |
job.control.total |
QCEW-based job controls |
employment.types |
RET,OFF,IND,OTH |
school.choice.coef.k8, .hs, .college |
input/school_location_coefficients_{k8,hs,college}.csv |
school.seat.release |
true (seat released when a student leaves school or the population) |
| Key | Calibrated value |
|---|---|
demographics.local.birth.rate.adjuster |
2.1724 |
demographics.local.marriage.rate.adjuster |
1.3436 |
demographics.local.divorce.rate.adjuster |
0.6867 |
demographics.local.leave.parents.adjuster |
2.7775 |
move.or.not.binomial.log.shift.parameter |
1.1114 |
construct.dwelling.demand.scaler |
0.2871 |
marriage.partner.distance.decay |
0.0462 |
population.control.total |
region (county controls) |
The levers are kept in landuse/calibration/calibrated_parameters.json.
Activities (MAST)¶
Written by activity/write_properties.py <name> <silo_scenario> <year> [sample] [skims]; the switches of the
coefficient set (activity/parameters/run_params.csv) are appended.
| Key | Value | Meaning |
|---|---|---|
tasha.generation |
true |
survey-distribution generation and scheduling |
tasha.dir |
coef/tasha |
distributions, scheduler parameters, return-home, outing and transit-leg models |
dest.choice.model |
mnl |
estimated destination and stop models |
dest.mnl.file |
dest_choice_mnl.csv |
their coefficients |
mode.choice.model |
tour_nl |
tour mode choice, nested or multinomial |
mc.coef.file |
tour_mode_choice.csv |
its coefficients by segment |
mc.vot.share |
0.5 | value of time = 0.5 × hourly household income |
mc.car.allocation |
tasha |
household car allocation |
mc.joint.rule, mc.joint.model |
representative, group |
joint-tour mode |
mc.park.share.file |
parking_paid_share.csv |
paid-parking share by area |
workplace.silo |
true |
workplaces from the land-use model |
sample.fraction |
0.1 | share of households simulated |
number.of.threads |
6 | |
auto.toll |
0 |
The seed is set with -Dabit.seed=N.
Assignment (MATSim)¶
The runner builds its configuration in code; each run writes the resolved output_config.xml.
java -Xmx32g -Dseed=1 -Dthreads=12 -Dincome.vot.form=usdot -cp baltimore-matsim-1.0.jar \
de.umd.matsim.RunBaltimoreToll <network> <plans> <schedule> <vehicles> <outDir> <iterations> <flowCap> [storageCap] [tollFile|NONE]
| System property | Default | Meaning |
|---|---|---|
seed, threads |
4711, 8 | random seed, threads |
modechoice |
false | modes fixed from the activity model |
reroute.weight, tam.weight, tam.range |
0.15, 0.10, 1800 s | re-routing and departure-time mutation |
income.vot.form |
power (usdot in the runs) |
value of time per person |
vot.share, vot.hours, vot.income.floor |
0.5, 2080, 5000 | value of time = 0.5 × income / 2,080, income at least $5,000 |
asc.car, asc.pt, asc.ride, asc.walk, asc.bike |
mode constants of the scoring | |
walk.maxdist, bike.maxdist |
2000 m, 5000 m | |
trucks |
false | optional truck layer |
Transit settings of the final runs: transfer penalty -Dpt.lineswitch=-0.13 (about 10 in-vehicle minutes) and
transfer walks up to -Dpt.beeline=300 m, set by pipeline/run_chain.py with MATSIM_PT=berlin MATSIM_PT_LIT=1.
Chain settings: flow capacity = sample (0.1), storage capacity = sample^0.75 (0.178), 100 iterations.