Author:

Keywords:

Earth observation satellites, Uncertainties of clouds, Expectation model, Chance constraint programming, Branch-and-price, Sample approximation, Column generation heuristic

Abstract:

This paper investigates the scheduling of multiple earth observation satellites (EOSs) under uncertainties of clouds. Firstly, we formulate the presence of clouds as stochastic events, transforming the problem into a stochastic programming problem. Based on different perspectives, we model the problem mathematically using both an expectation model and a chance constrained programming (CCP) model. Afterwards, for the first time, we employ a Dantzig-Wolfe decomposition and a column generation technique for the uncertain scheduling of EOSs. With respect to the expectation model, we devise a branch-and-price algorithm to solve the model optimally and efficiently. On the other hand, we first reformulate the CCP model as a mixed integer programming (MIP) model using sample approximation. Subsequently, considering the difficulties and the infeasibility of the branch-and-price algorithm for this MIP model, we suggest a column generation based heuristic algorithm to get \good" feasible solutions. By numerous simulation experiments, we verify the effectiveness and test the performance of our proposed formulations and approaches.