Rounding Error Analysis of an Orbital Collision Probability Evaluation Algorithm

We present an error analysis of an algorithm due to Serra et al. (Journal of Guidance Control and Dynamics, 2016) for computing the orbital collision probability in the short term encounter model. The algorithm reduces the numerical computation of the collision probability to that of the sum of a series whose coefficients are produced by a linear recurrence relation, and is specifically designed to avoid cancellation issues in the evaluation of the sum. While its numerical stability was observed experimentally and a truncation error bound was derived, the evaluation error was not studied. Here we give a rigorous bound on the accumulated rounding error when Serra et al.’s algorithm is implemented in floating-point arithmetic. For a unit roundoff u and a truncation order N, the bound is of the form (N + A)u + o(u), where o(u) stands for small compared to u, explicitly bounded terms. The constant A explicitly depends on the problem parameters and dominates N in practice. Our analysis is based on the observation that the generating series of the errors affecting each individual term is solution to a perturbed form of a differential equation satisfied by the Laplace transform of a function related to the collision probability.