The Smart Intersection: A Solution to Early-Stage Vehicle-to-Everything Deployment

Abstract

Connected vehicles and connected infrastructure technology are being deployed globally. Although the connectivity technology has a demonstrated capability of aiding collision prevention, the safety benefits are not maximized because of the limited coverage during the early deployment phase. Honda’s “smart intersection” research has led the company to a pilot deployment utilizing a state-of-the-art artificial intelligence solution in infrastructure sensing technology. This sensing technology employs a vision processing system to detect and differentiate nonconnected road users, including automobiles, emergency vehicles, motorcycles, and pedestrians. The detected object types are packaged into standardized wireless vehicle-to-vehicle (V2V) safety messages. These messages are transmitted from the roadside infrastructure to nearby connected vehicles. This article first discusses a few challenges associated with infrastructure sensing and collective perception and introduces techniques that ensure the infrastructure-generated V2V data transmissions are delivered in a real-time manner and with high message accuracy. We then examine the solutions through actual implementation accuracy assessments.