Automotive Doppler sensing: The Doppler profile with machine learning in vehicle-to-vehicle networks for road safety

Abstract

In this work, we present the fundamentals of a new sensing technique in Vehicle-to-Vehicle networks (V2V) called: Automotive Doppler Sensing (ADS), for providing road safety to connected drivers and connected autonomous vehicles by observing the Doppler profile. The Doppler profile displays the Doppler energy in the form of a high-resolution spectrogram which captures the non-line-of-sight (NLOS) reflections between stations. The Doppler profile can be analyzed in real-time for identifying vehicle dynamics as vehicles maneuver relative to each other. When machine learning is employed, the Doppler profile becomes a powerful new 360° “sensor” to provide both context awareness of the driving scenario and collision avoidance services. Experimental captures using real devices showcase the ADS capability to provide reliable awareness with high accuracy and few misclassifications in both line-of-sight (highways and surface streets) and NLOS (intersections) conditions.