Skip to content

SILO engine changes

MILES runs on SILO (msmobility/silo), commit 3066ca9. Its changes are shipped as landuse/engine/silo_changes.patch (modified files) and landuse/engine/src/ (new files at their package paths).

Classes

Class Module Status Role
ReslocModel useCases/maryland changed residential location; with resloc.twolevel=true it uses the two-level model
ReslocTwoLevel useCases/maryland new two-level model: PUMA (market size, own-race share, vacancy rule), then dwelling among 30 vacant dwellings
ReslocProperties useCases/maryland changed reads the resloc.* keys
Utility and verifier classes useCases/maryland new utilities of the estimated residential and car-ownership models, shared by engine and verifier
CarOwnershipModelMd useCases/maryland changed 0/1/2/3+ car logit with coefficients from file
SchoolDataMd useCases/maryland changed school location with the estimated K–8, high-school and college models; seat release
JobDataManagerImpl siloCore changed vacant-job workplace choice with the estimated model
JobDataProperties siloCore changed job.choice.*, job.search.impedance.file

Main class: de.tum.bgu.msm.run.SiloMstm.

Build and run

mvn compile -pl siloCore,extensions/schools,useCases/maryland
bash landuse/run/run_silo.sh landuse/run/<scenario>.properties 28g

run_silo.sh uses JDK 21, puts the compiled modules and landuse/run/classpath.txt on the class path and logs to run/logs/<scenario>.log.

From estimates to engine files

Script Writes
estimation/development/residential_location/ (final step) residential_puma_coefficients.csv, residential_dwelling_coefficients.csv and the PUMA and zone files
coefficients/export_car_ownership.py car_ownership_estimated_coefficients.csv, car_ownership_county_constants.csv
estimation/development/workplace_choice/w13_*.py workplace_choice_coefficients.csv
estimation/development/other_locations/l11_export_silo.py school_location_coefficients_{k8,hs,college}.csv and the school data files
coefficients/write_zone_attributes.py zone attributes

Checks: landuse/checks/check_*.py recompute utilities in the engine and compare them with \(X\beta\) from the estimation; the engine verifiers are switched on with system properties (see landuse/engine/README.md).

Adding a model

  1. Estimate it and write its coefficients to a CSV in landuse/engine_inputs/.
  2. Read the file in a properties class and add a switch key.
  3. Implement the utility in a pure class, so the engine and a verifier share it.
  4. Add a unit check that compares engine utilities with the estimation.