Towards C-V2X Enabled Collaborative Autonomous Driving
Yuankai He, Baofu Wu, Zheng Dong, Jian Wan, Weisong Shi
Abstract
Intelligent vehicles, including autonomous vehicles and vehicles equipped with ADAS systems, are single-agent systems that navigate solely on the information collected by themselves. However, despite rapid advancements in hardware and algorithms, many accidents still occur due to the limited sensing coverage from a single-agent perception angle. These tragedies raise a critical question of whether single-agent autonomous driving is safe. Preliminary investigations on this safety issue led us to create a C-V2X-enabled collaborative autonomous driving framework (CCAD) to observe the driving circumstance from multiple perception angles. Our framework uses C-V2X technology to connect infrastructure with vehicles and vehicles with vehicles to transmit safety-critical information and to add safety redundancies. By enabling these communication channels, we connect previously independent single-agent vehicles and existing infrastructure. This paper presents a prototype of our CCAD framework with RSU and OBU as communication devices and an edge-computing device for data processing. We also present a case study of successfully implementing an infrastructure-based collaborative lane-keeping with the CCAD framework. Our case study evaluations demonstrate that the CCAD framework can transmit, in real-time, personalized lane-keeping guidance information when the vehicle cannot find the lanes. The evaluations also indicate that the CCAD framework can drastically improve the safety of single-agent intelligent vehicles and open the doors to many more collaborative autonomous driving applications.