Vision-based formation control of aerial robots in the presence of sensor failure

Abstract

This article presents vision-based formation flight control for aerial robots with a special focus on failure conditions in visual communication. Then, by proposing and combining two strategies, a new solution is presented for formation control. In vision-based formation flight, the state variables of the leader are estimated using image processing and unscented Kalman filter. The follower adjusts its position with respect to the leader based on the results of the estimation. In the case of visual communication failure an error will occur in the estimation of variables, which would increase with the decreased image quality. In the first proposed strategy, during the failure emergence, the position of the follower aerial robot is obtained by combining the unscented Kalman filter's estimated velocity vector and the velocity vector before failure. The weighting coefficient of each velocity vector is obtained by fuzzy logic and based on image quality. In the second strategy, to reduce the possibility of collision between the members, the geometry of the formation pattern is expanded as a function of image quality and the distance between the members. The expansion coefficient is also extracted by a fuzzy inference method, and the desired distance between the members is increased as a function of expansion coefficient. These two strategies are combined to be used during failure periods. Finally, simulation studies are presented which are conducted based on the system nonlinear equations, a model with 6 degrees of freedom for each member, and the proposed visual noise model. Obtained results reveal the proper capability of the proposed hybrid strategy in terms of controlling the formation flight during failure conditions.