An intelligent visual servo control system for quadrotors

While quadrotors are becoming more popular, control of these unmanned air vehicles should be improved. In this study, a new intelligent image-based visual servo control system is proposed for the flight guidance control of quadrotors. Features are essential for visual servoing and the proposed system utilizes the features of a shape that provide a clear sight of the landing site instead of point features. Furthermore, the system focuses on three problems of visual servo control: finding an appropriate gain value under velocity limits, keeping the shape features in the field of view and tracking a moving target. As a solution to the first problem, a fuzzy logic unit that uses feature error and error derivative norms as inputs are deployed to assign the gain, adaptively. The second problem is solved by defining safe and risky regions in the image plane to take precautions before field of view missing. Another fuzzy logic unit is activated when the shape passes through a risky region to provide counter velocity in x or y direction and to drag the shape through the safe region. As the last stage, Kalman filtering with Potter's square root update is added to the proposed system to increase the feature tracking performance. This update also promises divergence avoidance. To show the performance of the proposed system, simulation results for fixed and moving targets under feature disturbance are presented for a quadrotor. The results verify that the proposed system is capable of handling visual servoing problems.