Quadrotor attitude estimation with data losses

This paper deals with the attitude estimation of a quadrotor in the presence of measurement losses. This situation may appear when the quadrotor is controlled through a network. The attitude estimation is a prerequisite to the implementation of any kind of control law. Moreover, the attitude cannot be directly measured and different sensor modalities must be used. The observer implemented is an Extended Kalman Filter derived from the Kalman Filter previously proposed by Sinopoli et al. [8]. The loss of data is modeled with a random process. The lost measurement is replaced with zero, the corresponding variance being set to an arbitrarily large value. All the scenarios that have been tested show the capability of the filter to properly estimate the quadrotor attitude even in the presence of a high rate of data loss. Experiments have been performed with a Matlab/Simulink simulator of the quadrotor developed at the GIPSA-Lab laboratory.