Adaptive on-ramp merging strategy under imperfect communication performance

Abstract

On-ramp merging is one of the important V2X (Vehicle-to-Everything) applications and is critical for both driving safety and traffic efficiency. The ramp vehicle needs to get information about the surrounding vehicles through V2X communications, and calculate the control strategy. But the communication performance issues of connected autonomous vehicles, such as communication delay and packet loss, will lead to instability and poor trajectory tracking accuracy of the control system, which can even affect the safety and efficiency of vehicle driving. However, vehicle communication performance issues in on-ramp merging scenarios are rarely taken into account. In this paper, we first analyze the delay and packet dropout problems in V2V (Vehicle-to-Vehicle) communication. Then, we propose a metric to measure the channel performance in the on-ramp environment called Exponentially Weighted Average Age-of-Information (E-AoI). And a reinforcement learning based algorithm for on-ramp merging control adaptive to communication performance is proposed. Lastly, we conduct simulation experiments with the traffic software SUMO (Simulation of Urban Mobility) and the V2V channel model to evaluate the proposed algorithm. Simulation results show that in the process of on-ramp merging in which the communication performance is considered, vehicle safety, comfort, and efficiency are guaranteed effectively by our proposed algorithm. In the case of persistent packet loss and delay due to channel congestion, our algorithm improves the security and efficiency of on-ramp merging compared to the basic DDPG (Deep Deterministic Policy Gradient) algorithm and the DDPG algorithm combined with MPC (Model Predictive Control) security constraints.