Enhancing Reliability in Infrastructure-Based Collective Perception: A Dual-Channel Hybrid Delivery Approach With Real-Time Monitoring

Standalone autonomous vehicles primarily rely on their onboard sensors and may have blindspots or limited situational awareness in complex or dynamic traffic scenarios, leading to difficulties inmaking safe decisions. Collective perception enables connected autonomous vehicles (CAVs) to overcomethe limitations of standalone autonomous vehicles by sharing sensory information with nearby road users.However, unfavorable conditions of the wireless communication medium it uses can lead to limited reliabilityand reduced quality of service. In this paper, we propose methods for increasing the reliability of collectiveperception through real-time packet delivery rate monitoring and a dual-channel hybrid delivery approach.We have implemented AutowareV2X, a vehicle-to-everything (V2X) communication module integrated intothe autonomous driving (AD) software Autoware. AutowareV2X provides connectivity to the AD stack,enabling end-to-end (E2E) experimentation and evaluation of CAVs. The Collective Perception Service(CPS) was also implemented, allowing the transmission of Collective Perception Messages (CPMs). Ourproposed methods using AutowareV2X were evaluated using actual hardware and vehicles in real-life fieldtests. Results have indicated that the E2E network latency of the perception information sent is around 30ms,and the AD software can use shared object data to conduct collision avoidance maneuvers. The dual-channeldelivery of CPMs enabled the CAV to dynamically select the best CPM from CPMs received from differentlinks, depending on the freshness of their information. This enabled the reliable transmission of CPMs evenwhen there was significant packet loss on one of the transmitting channels.