v1.0.1 Change The Objective Function

• changes:

  • change the measurement of a schedule from the total delay to the increased delay caused by new orders (a new function ScoreVtPairWithIncreasedDelay).
  • therefore, the objective function is changed to maximize the number of orders served (primary) and minimize the increased delay brought by serving new orders (the service rate is higher with this new objective function).
  • separate the scoring schedule step to an independent function ScoreVtPairsWithNumOfOrdersAndIncreasedDelay, so that ILP assignment is decoupled from the objection function.

• the data files used are same as v1.0.

v1.0 First Release With Data Files.

• Included:
  • data (included in datalog-gitignore.zip):
    • manhattan map, consisting of 4,091 nodes and 9,452 edges.
    • computed shortest path tables (path, travel time and travel distance), used as look-up table routing.
    • manhattan taxi trip data, extracted from three Wednesdays: 11th, 18th and 25th May in 2016. These have similar characteristics and are synthesized to four simulation scenarios of varying number of requests: 400k, 600k, 800k and 1000k. (attached below)
  • a simulation platform, consisting of a demand generator and a router.
  • three representative vehicle dispatch (order matching) algorithms: GI, SBA and OSP
  • two simple idle vehicle rebalancing algorithms: RVS and NPO
  • a python based video rendering tool to generate animations (need the background image in media-gitignore.zip)
  • external library: gurobi (need manual installation), fmt and yaml (automatically installed by CmakeLists)