A deep learning architecture to forecast space environment fragility and resilience

This paper introduces some novel metrics to quantify the fragility and resilience of the space environment. These metrics are derived from the percolation dynamics of an equivalent complex network modelling the evolution of the space environment. The network model represents groups of objects with nodes and their relationships with links. The idea is to study the time evolution of the relationships among groups of objects and derive an understanding of the global health of the space environment from the topological and dynamical properties of the network. Forecasting the future health of the space environment requires running long term propagations of the whole network dynamics. Therefore, this paper proposes a deep learning approach to forecast the long-term evolution of the network-theoretical metrics. This approach allows for the rapid assessment of different launch traffic models, active debris removal strategies or post mission disposal policies. The paper will present some representative examples, starting from the current population of resident objects in Low Earth Orbit.