AI and 5G for CCAM-The case of a roundabout

This paper presents a comprehensive study on the integration of Cooperative Connected and Automated Vehicles (CCAVs) in urban roundabouts. The research aims to assess the impact of CCAVs on traffic flow, safety, and driver comfort in a mixed-traffic environment. The state-of-the-art DiM400 Dynamic Driving Simulator of DriSMi Laboratory is employed, allowing the presence of a human-in-the-loop. It is coupled with SUMO for traffic simulation and VI-WorldSim for graphical representation. The study is part of the AI@EDGE project. It employs its Connected-Compute Platform (CCP) to develop an Edge Computing testbed and a new communication protocol called V2N2V. This protocol utilises a 5G network to perform all vehicular communications. A DRL policy is used to drive all CCAVs in the network. Its objective is to reduce vehicles’ fuel consumption and limit accelerations and jerks, optimizing comfort. It is placed in an Edge Node in the centre of the roundabout, receiving data from all vehicles and sending actions to the CCAVs. Preliminary tests involving human participants were conducted to evaluate their perception of CCAVs’ safety and efficiency. Scenarios with varying CCAV Market Penetration Rates (MPR) are considered, 20% and 80%, respectively. Results highlight an improvement in traffic flow and safety with increased CCAV integration. Participants, through a subjective questionnaire, expressed a preference towards the scenario with a higher MPR of CCAVs. They also rated the CCAVs safer with respect to Human- Driven (HD) vehicles. This suggests a positive reception towards automation in the roundabout.