Observability analysis of relative attitude determination algorithm based on stereo vision

To solve the problem that stereo-vision-based relative attitude determination algorithm couldn't correctly estimate gyro errors, the observability of each state in this algorithm is studied. Firstly, the filter model is designed based on the relative attitude error equation and stereo-vision measurement equation. Secondly, the unobservable subspace of the system is determined by choosing different variables as error model's states. Furthermore, the analysis is made on system structure decomposition, which reveals that this algorithm can only determine the relative attitude and gyro error's linear combination between the master star and the slave one, and when the feature points measured by stereo vision system are greater than two, increasing the number of feature points gives little help for improving the observability of gyro error. The simulation result validates the former discussion.