A Search Algorithm for Stochastic Optimization in Initial Orbit Determination

Optical observations constitute a source of angular measurements of a satellite pass. Commonly, these observations have short durations with respect to the satellite orbit period. As a consequence, the use of classical orbit determination algorithms, as Laplace, Gauss or Escobal methods, give very poor results. This thesis faces with the problem of estimating the orbital parameters of an orbiting object using its optical streak acquired by a telescope or a high accuracy camera. In this thesis a new technique is developed for the Initial Orbit Determination from optical data by exploiting the genetic algorithms. The algorithm works without restrictions on the observer location. A recent challenging problem is the Initial Orbit Determination with space- based observations. This thesis focuses on the problem of determinating the orbital parameters of a satellite from an orbiting observer in LEO, using short time observations. We present the results based on a simulation with the observer on a sun-synchronous orbit with a single observation of just 60 s. Monte Carlo simulations are presented with di erent levels of sensor accuracy to show the reliability of the algorithm. The algorithm is able to yield a candidate solution for each observation. The coplanar case is analyzed and discussed as well. Several test show the reliability of the algorithm varying the number of the observations, the initialization method, the observation period and the noise seed.