Multi-Objective Evolutionary formulations for design of hybrid Earth observing constellations

Abstract

With the recent advances in satellite miniaturization, communication and information technologies, and the advent of affordable small satellite launch services, there has been a paradigm shift in space exploration missions involving the transition from monolithic architectures formed by a large satellite to the concept of Distributed Spacecraft Missions (DSM), which fly multiple simpler and less costly satellites to offer increased capabilities such as better temporal, spatial, and angular sampling. Despite the potential to provide higher science return and novel data products, the orbit selection problem for Earth Observation DSMs requires a much more complex constellation design process, which involves several interrelated design variables and conflicting objectives. Because of this, prior work has focused mostly on relatively simple DSM architectures consisting of a single type of constellation, such as homogeneous Walker constellations. This paper presents a novel evolutionary formulation (i.e., chromosome and operators) in the context of Multi-Objective Evolutionary Algorithms (MOEA) that allows for the exploration of large tradespaces of non-Walker hybrid satellite constellations with diversity of orbital parameters. The idea behind this new formulation is that it allows to search through the space of different types of constellations – such as Walker formations, Sun-synchronous trains and string-of-pearls among others – and combinations thereof. This type of hybrid constellations have not been studied in detail. The methodology presented in this paper helps overcome the combinatorial explosion resulting when opening up the design space to include non-symmetrical configurations and relaxing constraints in the values of some orbital parameters (e.g. choosing a common altitude or inclination for all satellites forming the constellation). The proposed formulation is compared with a state-of-the-art evolutionary formulation using a variable-length chromosome in 5 different problems including the observation of symmetrical, asymmetrical, connected and disconnected regions of interest. Results show that the proposed formulations achieve better convergence and convergence rate than the state-of-the-art. The proposed method can reduce the effort required to design a problem formulation for each problem instance, while also reducing the risk of missing potentially good architectures due to formulations that are too restrictive and rely too much on previous experience and expertise.