Skip to content

Integrate Jiaoyang's Scheduler into the System

Zi Liang edited this page Mar 20, 2019 · 4 revisions

1. Info of Jiaoyang's scheduler

1.1 Input

The input includes the last node and time that each aircraft just passed by, as well as the scenarios.

1.2 Output

The output is a priority list of aircrafts and their itineraries (only contains nodes in model).

1.3 Assumptions

  1. All of the scenarios within the next scheduling period is known by the scheduler.
  2. The simulator can avoid conflicts with pilot-based model.

1.4 Main logic

2. Differences from the current simulator

  1. The simulator is in Python, and has a relatively complex logic. Jiaoyang's scheduler (called the new scheduler later) is in C++. Possible solution: since communication via files might cause problems in heterogenous environments and slow down the simulation, it would be better if we can rewrite the new scheduler in Python if its logic is not very complicated.

  2. The simulator periodically get itineraries from the scheduler, while the new scheduler later makes schedule only once. Possible solution: The new scheduler can be changed to make periodical schedules.

  3. Current scheduler in the simulator only schedules for the ready aircraft. If an new aircraft is added into the system between two scheduling times, it will ask the scheduler for an itinerary whenever it is ready. The new scheduler only schedules at scheduling time, and it takes into consideration all aircraft ready/will ready before the next scheduling. Possible solution: store the itinerary of future aircrafts in the scheduler, and returns it back when it is ready and asks for an itinerary.

  4. Currently the new scheduler haven't run on SFO node-link model yet. Possible solution: Zi exports the routing table (get the shortest path between nodes) in the RoutingExpert into json by this week for Jiaoyang to test the new scheduler.

  5. The simulator provides the position of aircraft as a precise point in the link, which is usually different from the nodes in the node-link model.

Acknowledgements

Thanks for Zhongyi's help and suggestions.