Proposal

Main paper can be found here.

  • Focuses on DAYOPS (Day of Operations Scheduling) problem in real-time under minor perturbations from the original plan.
  • Both activity start and duration times are considered as variables
  • Authors show that the dual of the resulting model is a network problem and thus has a linear complexity of time to solution

Summary

  • DAYOPS involves determining real-time changes to flight schedules when perturbations occur to minimize customer inconvenience and airline costs
  • planning process of large airlines involves the following 4 phases:
    • flight scheduling
    • fleet assignment
    • aircraft routing
    • crew scheduling
  • these phases are done offline with lots of data and time at hand. But for DAYOPS scenario both are scarce
  • they show that while the primal equation with constraints scales polynomially in runtime wrt the model size, the dual form only scales linearly. Thereby making the latter a suitable candidate for realtime requirements
  • Authors use CPLEX linear optimizer to solve the resulting system of constrainted equations