TY - GEN
T1 - Adaptive dynamic programming for airport deicing
AU - Zhang, Zirun
AU - Rosenberger, Jay M.
AU - Chen, Victoria C.P.
AU - Zeng, Li
AU - Bergman, Dan
N1 - Publication details (e.g. title, author(s), publication statuses and dates) are captured on an “AS IS” and “AS AVAILABLE” basis at the time of record harvesting from the data source. Suggestions for further amendments or supplementary information can be sent to [email protected].
PY - 2013
Y1 - 2013
N2 - An adaptive dynamic programming (ADP) approach for assigning aircraft to deicing pad locations at the Dallas-Fort Worth (DFW) International Airport is developed to minimize the impact on dissolved oxygen in airport receiving waters. Lack of dissolved oxygen killed fish at DFW in 1999. The DFW deicing activities ADP problem is stochastic, finite-horizon, non-stationary, and non-convex with integer decision variables and a very high-dimensional (near) continuous state space. Using treed regression functions to approximate state transitions and predict dissolved oxygen at various water sites, a mixed integer linear programming problem is formulated within each stage to maintain dissolved oxygen (DO) in current and future stages. A waiting time penalty is also modeled in the stage-wise objective function. Uncertainties exist in deicing activities because the airport only controls the location of deicing, while the pilot controls the actual deicing activity. Distributions for these uncertainties are empirically-derived from actual airport data. Because treed regression models are non-convex, non-convex value approximation methods are used and another mixed integer linear programming problem is developed. This paper focuses on formulating and solving the mixed integer linear programming problem of last stage after the framework for solving the entire deicing ADP problem is presented.
AB - An adaptive dynamic programming (ADP) approach for assigning aircraft to deicing pad locations at the Dallas-Fort Worth (DFW) International Airport is developed to minimize the impact on dissolved oxygen in airport receiving waters. Lack of dissolved oxygen killed fish at DFW in 1999. The DFW deicing activities ADP problem is stochastic, finite-horizon, non-stationary, and non-convex with integer decision variables and a very high-dimensional (near) continuous state space. Using treed regression functions to approximate state transitions and predict dissolved oxygen at various water sites, a mixed integer linear programming problem is formulated within each stage to maintain dissolved oxygen (DO) in current and future stages. A waiting time penalty is also modeled in the stage-wise objective function. Uncertainties exist in deicing activities because the airport only controls the location of deicing, while the pilot controls the actual deicing activity. Distributions for these uncertainties are empirically-derived from actual airport data. Because treed regression models are non-convex, non-convex value approximation methods are used and another mixed integer linear programming problem is developed. This paper focuses on formulating and solving the mixed integer linear programming problem of last stage after the framework for solving the entire deicing ADP problem is presented.
KW - Adaptive dynamic programming
KW - Deicing/anti-icing activities
KW - Mixed integer linear programming
KW - Non-convex approximation
KW - Non-convex optimization
UR - https://www.scopus.com/pages/publications/84900309205
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-84900309205&origin=recordpage
M3 - RGC 32 - Refereed conference paper (with host publication)
T3 - IIE Annual Conference and Expo 2013
SP - 3670
EP - 3678
BT - IIE Annual Conference and Expo 2013
PB - Institute of Industrial Engineers
T2 - IIE Annual Conference and Expo 2013
Y2 - 18 May 2013 through 22 May 2013
ER -