Skip to content
/ GDP Public

Markov Decision Process for Ground Delay Program planning

Notifications You must be signed in to change notification settings

sisl/GDP

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GDP

This directory contains a Markov Decision Process model for the Single Airport Ground Holding Problem (SAGHP) and three solution approaches- one based on approximate dynamic programming, the Ball et al. model, and the Richetta-Odoni dynamic model. This code is intended as a supplement to the article, Ground Delay Program Planning using Markov Decision Processes by Jonathan Cox and Mykel J. Kochenderfer, published in the Journal of Aerospace Information Systems.

Run the models

The files in the examples directory simulate each model for one day at a single airport. The output is the sum of rewards for one simulation using landing capacities randomly generated by the AAR Distribution Prediction Model (ADPM). The choice of airport, delay costs, and model-specific parameters can be set in adp.jl, ball.jl, and ro.jl.

To run a simulation, cd to the examples directory and execute a file. The file adp.jl uses approximate dynamic programming to choose actions, the file ball.jl uses the Ball et al. model, and the file adp.jl uses the Richetta-Odoni dynamic model. The simulation may take several minutes to run. When the simulation completes, the sum of rewards will be printed to the screen.


$ cd examples
$ julia adp.jl

Sum of rewards for Approximate Dynamic Programming at EWR
-15.0

Layout

examples/
    adp.jl                              simulates using approximate dynamic programming 
    ball.jl                             simulates using the Ball et al. model
    ro.jl                               simulates using the Richetta-Odoni dynamic model
    
mdp/
    MDP.jl                              module for simulating a Markov Decision Process
    GDP.jl                              MDP model

solvers/
    adp/
        ADP.jl                          approximate dynamic programming module
    ball/
        Ball.jl                         the Ball et al. model
        BallSeq.jl                      module for running the Ball et al. model iteratively
    ro/
        RO.jl                           the Richetta-Odoni dynamic model
        ROSeq.jl                        module for running the Richetta-Odoni dynamic model iteratively

adpm/
    readParams.jl                       extracts ADPM parameters 
    predictAAR.jl                       generates landing capacities
    params/                             ADPM parameters
        adpm_params_ewr
        adpm_params_sfo

datatools/
    DataTools.jl                        functions for reading flight schedules and forecasts

data/                                   schedules, weather forecasts, and landing capacities
    aars_EWR.csv                        
    aars_SFO.csv                       
    aars_EWR_adp.csv                        
    aars_SFO_adp.csv                       
    aar_first_EWR.csv                 
    aar_first_SFO.csv                
    EWR_2013-05-02.csv                 
    SFO_2013-05-02.csv                
    forecast.jld                     

About

Markov Decision Process for Ground Delay Program planning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages