Symphotic Design of an Edge Detector for Autonomous Navigation

Abstract

Autonomous navigation systems rely on the collection and processing of large datasets with advanced and memory-intensive algorithms. The requirements in terms of computing power and energy consumption can be significant. In this work, we propose a passive electromagnetic structure that can reduce the computational load by detecting edges (obstacles) in the far field. This is accomplished by probing the scene and processing its scattering on the physical layer, at the speed of light in the medium. By using a recently developed inverse design method, we present an edge detector able to detect obstacles at 15 different locations with an average efficiency of 97% and minimal crosstalk. We call this class of devices that can integrate a vast number of distinct optical functions with high efficiency symphotic.