Design, Implementation, and Evaluation of a Roadside Cooperative Perception System

Abstract

Recently, with the advancement in autonomous driving, artificial intelligence, and vehicle-to-everything (V2X) communications, cooperative perception between roadway infrastructure and connected and automated vehicles via V2X communications has attracted increasing attention. However, most of the systems and prototypes reported are still at an initial stage. This paper introduces a newly developed and deployed roadside cooperative perception system with an edge-cloud structure and multiple kinds of sensors, including fisheye cameras, thermal cameras, and long-distance radars. The system is deployed at a roundabout at the intersection of State St. and W. Ellsworth Rd. in Ann Arbor, Michigan, U.S. With an edge-cloud structure, it can execute high-performance detection algorithms on edge devices at the field and collect data and analyze it on the cloud. The performance, in relation to detection rate, localization accuracy, and latency, is analyzed using the data collected from the field. These results show that the system has high potential for a wide range of applications, including infrastructure-vehicle cooperative perception, traffic data collection, and road safety studies. Two example applications—traffic volume monitoring and road safety warning—are introduced in the latter part of the paper to demonstrate the system’s capability.