Robust M-Type Error-State Kalman Filters for Attitude Estimation

State estimation techniques appear in a plethora of engineering fields. Both standard Kalman filter (KF) and its nonlinear extensions, as well as particle filters, consider a known system model (i.e., functions and noise statistics), an assumption which may not hold in practice. A problem of particular interest is how to deal with outliers in the observation model. A possible solution is to resort to the framework of robust statistics, where a robust score function is used to mitigate the impact of outlying measurements, leading to robust M-type KFs. In this contribution, some of these robust filtering results are extended to the case where states may live on a manifold (unit norm quaternion), and propose robust iterated error-state M-type KF solutions. An illustrative example is provided to show the performance of the proposed filter and support the discussion.