Interacting Multiple Model Filter Based Autonomous Landing Considering Camera Model Uncertainty

This paper deals with the state estimation and control for camera-assisted landing. Uncertainties in measurement models, such as lens distortion, can adversely affect the camera's target detection and navigational functions. We propose an effective state estimator based on the Interacting Multiple Model filter to account for the uncertainties in the camera measurement model. We further introduce a phase update rule, which determines the feasibility of autonomous landing in a stochastic manner, and filter and control management logic is proposed by which to handle the different phases. The proposed framework is simulated using a multi-rotor type UAV equipped with a camera and rangefinder in the Gazebo simulation environment. The experiments validate the performance of the framework in terms of accuracy in state estimation and control.