Item

Abstract

Flying insects use highly efficient visual strategies for stabilizing their motion in three-dimensional space. We present a flight control model that uses a combination of biologically inspired, visual feed-forward mechanisms for stabilizing attitude (i.e. pitch and roll angles) and altitude during translatory motion. The attitude sensor exploits the position invariant vertical intensity gradient that exists in most natural open environments to orient the upper surface of a flying agent towards the region of maximum brightness (dorsal light response). Altitude is controlled using distance information contained in the frontoventral translatory optic flow (motion parallax). Our results from open-loop computer simulations show that the signals produced by these mechanisms robustly indicate the direction of deviation from a certain attitude angle or ground distance. We argue that in a closed control loop, these qualitative signals can be sufficient for flight stabilization. We present closed-loop trajectories of a simulated agent equipped with both mechanisms, flying over a textured, uneven surface. The agent shows robust flight behavior with six degrees of freedom.