General framework for the reconfiguration of satellite constellations

From remote sensing to navigation and communication, satellite constellations have become an indispensible component of our society’s infrastructure. Recent events, including China’s intercept of their Feng Yun-1C weather satellite and the United States’ intercept of a non-functioning satellite, have dramatically increased the amount of space debris, which poses an increased risk for on-orbit collisions. When the loss or degradation of a satellite in a constellation is experienced—be it from a collision with space debris, on-orbit malfunctions, or natural causes—the constellation may no longer be capable of fulfilling its mission requirements. Instead of simply accepting the degraded performance, stakeholders may consider reconfiguration of the remaining spacecraft. In this research, a general framework for the reconfiguration of satellite constellations is developed. The key characteristic that separates this research from others that have come before it is that the future state of the reconfigured constellation is not assumed a priori; rather, it is this state that is found. In other words, the reconfigured geometry can occupy any feasible region (this is mission dependent) of the design space in an attempt to provide optimal performance with respect to multiple-criterion. These include the propellant expended, time of transfer, reduction in mission life, coverage performance, and risk due to maneuvering. When modeled as mathematical functions, some of these concerns exhibit continuous behavior; however, most have nonlinear, discrete, discontinuous, and/or multimodal characteristics. The framework adapts a best-in-class parallel Multi-Objective Evolutionary Algorithm to approximate the optimal hypervolumes for this complex tradeoff-space. Several loss scenarios for the Global Positioning System constellation are presented to demonstrate the framework. An a posteriori procedure for decision support is introduced that enables down-selection to a final design from thousands of non-dominated reconfiguration alternatives. Among the significant results to emerge from this research are the lessons learned as a result of the application of stochastic optimization to the constellation design problem domain. One such lesson indicates that objective functions related to coverage have the greatest influence on the multi-modality of the design space. Other results demonstrate that increasing the number of design variables and/or the application of operational constraints (such as fuel budgets) do not necessarily make the reconfiguration problem more difficult to solve; in some cases it becomes easier. The method itself is successful in providing a global context to decision makers that allows for defendable design selection in what was previously a computationally intractable optimization problem.