An investigation of the effects of model resolution on control of a quadrotor micro air vehicle

Abstract

Control laws derived from dynamic inversion are dependent on the model which describes the system behaviours. In control of autonomous systems, simple models which neglect rotor dynamics or non-linearities may be employed to develop linearising feedbacks for dynamic inversion control. The presence of these additional dynamics or non-linearities in the real system may produce undesirable effects when controlled by a dynamic inversion controller which does not consider them. Thus, multi-resolution modelling is employed to investigate the impact of unmodelled dynamics on a dynamic inversion controller derived from simple models. Multi-resolution modelling (MRM) is the practice of describing the same system at multiple levels of complexity, each level containing an increase in state or additional non-linearities with respect to that below it. MRM and dynamic inversion are applied to the quadrotor micro air vehicle (MAV) as a test case. Three models of the quadrotor, each comprising a level in the multi-resolution model and differing in complexity, are specified. Dynamic inversion controllers are then derived from each level and applied to plants of equivalent or higher level. It is found that increasing the desired closed-loop response can produce unstable behaviour in systems with dynamics not considered by the controller. Incorporating these dynamics by using a controller derived from a higher-resolution model is shown to produce a stable response at these bandwidths.