I have been fortunate to be engaged in research since my senior year at Tsinghua University in China. Over the past nine years, my research trajectory and publication record have passed through hydraulic engineering, geotechnical engineering and environmental engineering, before finally reaching transportation engineering.
In the future, I will continue the journey by focusing on transportation. My research interests though extend to other disciplines within CEE and beyond.
Transportation system
|
Data analysis
|
Air traffic management
| Airline economics
|
My dissertation introduces the goal of predictability into cost optimization of an air traffic management tool—Ground Delay Program (GDP)—under capacity uncertainty for a single airport case. This is accomplished by modifying traditional GDP delay cost functions so that they incorporate predictability. Extra premiums are assigned to unplanned delays, and planned but un-incurred delays, due to their unpredictability.
Two stochastic GDP models are developed based on deterministic queueing theory and continuous approximation to estimate the delay components in the cost functions: a static no-revision model and a dynamic model considering one GDP revision. GDP modeling is a complex problem because GDP decisions are made under capacity uncertainty and inbound flight traffic is coming from different distances. Another challenge is that air traffic managers commonly revise GDPs when weather forecasts are updated. This results in complicated dynamics since the impact of a revision on an individual flight depends on its schedule and flight time. Unlike more widely used mathematical programming approaches, our method does not require detailed inputs about flight schedules and capacity scenarios. The output from our models—the time when the planned arrival rate increases to reflect good weather conditions—is also much simpler. In this way, the sensitivity of the optimal planned capacity recovery time to different assumed cost functions can be easily compared.
The optimization results from the case study show that unpredictability clearly matters, particularly in the more realistic case where revision is allowed. Of the two unpredictability cost parameters, the one for unplanned delay has a stronger impact than the one for planned un-incurred delay. Depending on the values of unpredictability premiums, considerable cost may be saved if the decision takes predictability into account.
A 5-minute briefing of the work can be found in a press release by Informs. A complete story can be found in a Transportation Science paper titled incorporating predictability into cost optimization for ground delay programs.
