Feedback control of dynamical systems using neuromorphic vision sensors

Abstract

The recent development of neuromorphic vision sensors, which provide an asynchronous, high-speed alternative to conventional cameras has lead to a considerable amount of research into their applicability to robotic control systems. However, algorithms for onboard control of mobile robotic platforms such as automobiles or aircraft using these sensors are lacking and in fact almost all existing implementations keep the sensor stationary. This research has several objectives. First, to develop a rigorous understanding of how to use asynchronous temporal contrast vision sensors for heading regulation and tracking in such a way as to fully leverage the remarkable properties of these sensors including high bandwidth, low latency and low power consumption. Second, to provide a theoretical and experimental comparison between neuromorphic vision sensors and conventional cameras in the context of this problem. Finally, to describe and test algorithms for high-speed motion planning in cluttered environments using neuromorphic vision sensors.