An Evolutionary Multi-objective Approach for Stochastic Air Traffic Network Flow Optimization

The Stochastic Air Traffic Network Flow Optimization (SATNFO) problem aims to seek a set of optimum and robust flight plans to ensure a safe, orderly and expeditious air traffic flow in the presence of uncertainties. Due to the very natures of multi-objective, large-scale and non-separable in the SATNFO problem, this paper sparks an evolutionary multi-objective optimization way for solving it. Firstly, we formulate it as a multi-objective problem, with performance and robustness as separate goals. In this model, robustness, which indicates the ability of a flight plan to cope with negative effects of uncertainty, is quantified and introduced as an objective. And, two conflicting performance objectives, i.e., minimizing the workload as well as the flight delays over the network, are involved. Then, we present an adaptive metaheuristic algorithm, termed as aNSGA-II, to solve the SATNFO problem. In aNSGA-II, a parameter adaptive mechanism is designed to dynamically adjust the probability of crossover and mutation based on problem context and evolution mechanism. It helps to balance exploitation and exploration during the evolutionary process, and thus maintain diversity of solutions and improve the convergence performance of the algorithm. Empirical studies using real data of flights and network in China are carried out, and show ability of our approach in providing efficient and robust flight plans and supporting better decision-making for air traffic controllers in a stochastic scenario.