TA Code Documentation

Major Changes I Made

• Modified frank_wolfe_2.py, AoN_igraph.py, and frank_wolfe_heterogeneous.py to record, update, and return path flows
• The end of frank_wolfe_heterogeneous.py or the script run_cog_cost.py shows how I run the cognitive cost model:
  • Iterate over the app (routed) user percentage (alpha) from 0-100%
  • Load LA_net input files
  • Modify two links to increase capacity
  • Set threshold (1000) so that links under that capacity have increased cognitive cost (3000x) for non app (non routed) users
  • Using 50% demand and modify to maintain 4000 veh/hr units of flow
  • Get graph with adjusted cognitive costs for non app users
  • Split demand based on alpha
  • Run frank-wolfe algorithm (every iteration I also do a sanity check to make sure path flows and link flows add up)
  • Output link flows, path flows, and nash distance
• frank_wolfe_heterogeneous2.py is my implementation of the restricted path choice model
• graph_results.py is a script that
  • Compiles OD data from frank-wolfe output into a single file for one OD pair (all paths for that OD pair at all app usage percentages in decreasing path flow order)
  • Calculates nash distance if necessary (old versions of frank_wolfe_heterogeneous.py didn't do that)
  • Produces OD data used in static model dashboard
  • Graphs path flow and travel times against app usage percentage for app users and non app users
• convert_for_dash.py makes all files necessary for dashboard (including the OD data that the last script can also create)