A multi-arc approach for chaotic orbit determination problems

International audienceChaotic dynamical systems are characterized by the existence of a predictability horizon, connected to the notion of Lyapunov time, beyond which predictions of the state of the system are meaningless. In order to study the main features of orbit determination in presence of chaos, Spoto and Milani (2016) applied the classical least squares fit and differential correction algorithm to determine a chaotic orbit and a dynamical parameter of a simple discrete system-Chirikov standard map (cf. Chirikov (1979))-with observations distributed beyond the predictability horizon. They found a time limit beyond which numerical calculations are affected by numerical instability: the computability horizon. In this article we aim at pushing forward such inherent obstacle to numerical calculations in chaotic orbit determination by applying the classical and the constrained multi-arc method (cf. Alessi et al. (2012)) to the same dynamical system. These strategies entail the determination of an orbit when observations are grouped in separate observed arcs. For each arc a set of initial conditions is determined and, in the case of the constrained multi-arc method, all subsequent arcs are constrained to belong to the same trajectory. We show that the use of these techniques in place of the standard least squares method has significant advantages: not only can we perform accurate numerical calculations well beyond the computability horizon, in particular the constrained multi-arc strategy improves considerably the determination of the dynamical parameter