An optimal smoothing approach for trajectory reconstruction in planetary exploration

This paper deals with fusion of real-time and off-line measurements, formulated in terms of a nonlinear least-squares optimization problem, and solved analytically by linearizing the measurement model. The proposed approach is applied to a planetary exploration rover, equipped with an EKF based localization system using fixations to unknown fixed landmarks. The smoothing method enables fusing the off-line measurements, related to the landmarks, with the realtime EKF estimates. The analytic solution of the resulting optimization problem has major advantages over standard methods, avoiding risks of solution divergence and of convergence to local minima, and reducing the computational load. An analytic evaluation of the linearization error is presented, along with simulation results to demonstrate the effectiveness of the proposed approach.