Adaptive Control of Multistage Airport Departure Planning Process using Approximate Dynamic Programming

Many service enterprise systems such as the airport departure systems are typical multistage multivariable systems with non-linear complex interactions between stages. These systems function over a wide range of operating conditions and are subject to random disturbances, which further enhance the non-linear characteristics. Also, there are many uncertain factors which often makes it is difficult to describe the process dynamics with complete information and accurate physical and empirical models. Adaptive controllers based on the analytical and/or artificial intelligence techniques can provide improved dynamic performance of the multistage process by allowing the parameters of the controller to adjust as the operating conditions change, and are known to operate in model free environment. One such example of an adaptive controller, is the combination of analytical dynamic programming methods and artificial intelligence techniques to achieve superior control of operations and improved quality of finished products. This new branch of research has become known as Approximate Dynamic Programming (ADP) methods. This paper first presents a state-of-the-art review including the advantages and limitations of ADP methods. Next, it develops a novel multiresolution assisted reinforcement learning controller (MARLC) based on ADP principles, which is used in an agent-based control model for improving the performance quality of the multistage airport departure planning process. The research is ongoing in collaboration with the Center for Air Transportation Systems Research at George Mason University.