Abstract: Autonomous orbit determination for a two satellite system with modelling errors is concerned in this paper．The performance of the autonomous orbit determination system is greatly affected by these errors in the dynamic model．One way to deal with these errors is treating them as unknown biases and estimating them with dynamic states together. Based on k order local weak observability theory, sufficient and necessary conditions of k order local weak observability are mathematically proved for the orbit determination system with modelling errors. The Extended Kalman Filter is applied to estimate the orbital states and systematic errors simultaneously．Simulation results show that the estimation algorithm is effective when the modeling errors are constant or slow time varying vectors.

Key words: orbit determination, modeling error, error estimation, autonomous navigation, observability analysis, nonlinear systems