This paper conveys attitude and rate estimation without rate sensors by performing a critical comparison, validated by extensive simulations. The two dominant approaches to facilitate attitude estimation are based on stochastic and set-membership reasoning. The first one mostly utilizes the commonly known Gaussian-approximate filters, namely the EKF and UKF. Although more conservative, the latter seems to be more promising as it considers the inherent geometric characteristics of the underline compact state space and accounts-from first principles-for large model errors. The set-theoretic approach from a control point of view is addressed, and it is shown that it can overcome reported deficiencies of the Bayesian architectures related to this problem, leading to coordinate-free optimal filters. Lastly, as an example, a modified predictive filter is derived on the tangent bundle of the special orthogonal group(3).
INTRODUCTIONAttitude and rate estimation is an important aspect of aerial robotics. Throughout the decades, it has proven very accurate and versatile in applications from the first Low Earth Orbit (LEO) satellites [1] to unmanned aerial vehicles (UAVs) [2] and from the unmanned aerial systems [3] to recent aerial robotic workers [4]. At the same time, technological and technical advances allow for increased specifications of autonomy in conjunction with precise and agile manoeuvring. Consequently, position and orientation (attitude) control constitutes a field of research that is vital component of aerial robotics. In many cases, the model can be decoupled and attitude control can be implemented independently from position control [2]. Lately, more focus has been given to attitude controllers due to the increased difficulty and complexity of the specific control problem [5]; the success of these controllers relies upon the accurate knowledge of the real orientation and the angular rate of the aerial robot. Thus, it is imperative to develop efficient attitude filters, to deal not only with the measurement noise but also with the model errors.When a-priori statistical information is available, such uncertainties are represented by utilisation of the stochastic framework. Subsequently, model errors and measurement noise areThis is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.