Yash Chandramouli, Amazon Project Kuiper; Liberty Shockley, Amazon Project Kuiper
Keywords: collision probability, lifetime, debris, space safety
Abstract:
As space safety becomes paramount in spacecraft design, there is a growing need for precision and transparency in the available tools for analyzing the on-orbit and reentry risk of a given space mission. Programs like NASA’s Debris Assessment Software (DAS) or ESA’s Debris Risk Assessment and Mitigation Analysis (DRAMA) tool are commonly used to calculate critical space safety metrics such as satellite deorbit time, lifetime collision probability, and reentry human casualty risk as a part of the initial regulatory licensing process. However, these models are designed primarily as compliance assessment tools, with limited options to assess multiple spacecraft configurations simultaneously and with little visibility into internal computations. These factors make it prohibitively labor-intensive for spacecraft designers to explore large trade spaces or derive useful insights when considering mission space safety impacts during initial trade studies. Additionally, without insight into what design changes might affect the final computations, the space safety impact of a system is often only assessed at the time of licensing, not accounting for the impacts of mission and satellite architecture changes that might occur throughout the design lifetime of the spacecraft. With no analytical method of computing these metrics, a spacecraft mission designer who wishes to incorporate space safety metrics into their system trade space is forced to iteratively “guess-and-check” to design mission trajectories that comply with regulatory requirements.
Sensitivity studies on input parameters could provide a window into these complex models and give spacecraft mission designers the tools they need to continually and efficiently assess their compliance with space safety regulations throughout the satellite design process. Sensitivity analysis methods aim to capture the impact that input parameter variations have on the output of a model. These methods can be local or global. Local methods, such as one-at-a-time (OAT) methods, capture the sensitivities around a baseline input set, often by modifying a single variable at a time and tracking the corresponding changes in the output. Global methods instead aim to capture a broader picture of sensitivities across a wide range of datapoints. One global method is estimating the partial derivatives of a surrogate model mapping the output with respect to the input, often by extracting the coefficients from a regression model. Another commonly used global method is Sobol’s method that computes “sensitivity indices” which represent the proportion of the output variance that is captured by knowledge of a subset of input variables. While analytically difficult to solve for, these indices can be estimated to high accuracy through Monte Carlo analyses.
This paper presents an approach that uses global sensitivity analysis to help spacecraft mission designers more efficiently prioritize and explore the trade space around a given space safety metric. Specifically, this paper focuses on assessing the sensitivity of DAS outputs of lifetime large object collision probability. After generating a dataset of space mission profiles across LEO, we then compute Sobol indices with respect to inputs such as satellite semi-major axis, inclination, deorbit year, and area-to-mass ratio. We then discuss novel insights showed by this analysis. Lastly, we demonstrate how sensitivity analysis can be used by spacecraft mission designers to prioritize changes to their system, estimate the impact of variations in parameters, and compute equivalent changes to then be passed into broader system-level trade studies.
The final results report new DAS performance data that are valuable to spacecraft operators and regulators, and provide valuable insights and tools to spacecraft designers who wish to incorporate space safety concerns into their trade studies. The method presented can also be extended to other space safety model outputs such as satellite reentry casualty risk or small object collision probability.
Date of Conference: September 16-19, 2025
Track: Space Debris