Dynamic Analysis and Modeling of DNN-based Visual Servoing Systems

The integration of deep learning and control techniques has created robotic systems capable of implementing visual servoing and navigating autonomously in unknown environments. However, analyzing the effect that timing interactions between controllers and deep neural networks have, has received little attention in the literature. In this paper we describe a novel model that includes the effects that detection loss and inference latency have on the controller of a visual servoing system. To test our model we created a target tracking system consisting of a video camera mounted on a moving platform that tracks objects using deep neural networks.