Genetic Design of Target Orbits for a Temporary Reconnaissance Mission

R ECENTLY, there has been increasing demand for highresolution geospatial information from space for civilian, as well as military users. However, there are obstacles to realizing a constellation of commercial high-resolution satellites over Earth’s entire surface remains subject to restrictions; these obstacles are primarily caused by the development cost and such an operation itself. For this reason, a sparse coverage constellation focusing on discontinuous coverage over a local area (or target) of interest could be an alternative. Several studies addressing the sparse coverage constellation design problems have been conducted by Lang [1], Crossley and Williams [2], and Williams et al. [3]. Ferringer and Spencer [4] designed a sparse constellation to resolve the conflict between revisit time and resolution. Meanwhile, a new constellation design proposed by Schiff and Mailhe [5] was applied to use resources from previously or soon-to-be launched satellites. Recently, a natural orbit whereby all sites are visited within a time frame without maneuvering was introduced by Abdelkhalik and Mortari [6]. However, very little of the work offered appropriate solutions for a “temporary reconnaissance mission” involving a few low-Earthorbit (LEO) satellites over a particular target site during a specified time. This is because each LEO satellite is currently in its own orbit with a limited fuel budget for orbital maneuvering and the sensor characteristics of each satellite are not identical. In this Note, the authors propose a new approach to find the target orbits of each satellite to establish a temporary reconnaissance constellation mission to minimize the average revisit time (ART) while satisfying the constraint on fuel limit. To achieve this goal, we employed a genetic algorithm (GA) to handle the discontinuity and nondifferentiability of the objective function for the given problem. The performance of the GA tends to be problem-dependent. Thus, preliminary efforts to improve the fitness function in dealing with a GA value are also presented.