Prediction and Control of Connected Mixed Traffic under Different Information Flow Topologies

Different platoon controls of connected automated vehicles have been studied to improve the entire fleet's overall energy efficiency and driving safety. The platoons can be used during highway cruising to reduce unnecessary braking, shorten required headway, and thus improve traffic capacity and fuel economy. They can also be used in urban driving to improve traffic efficiency at intersections. However, there remain two problems that prevent the technology from achieving maximum benefit. First, the presence of human-driven vehicles will change the behavior of the fleet and platoon control of connected mixed traffic. Second, the communication uncertainties impose negative impacts on the dynamics of the platoon. A high-performance state predictor for surrounding vehicles can reduce the human-driven vehicle's influence and help handle communication uncertainties better. This article proposes a novel inverse model predictive control (IMPC)-based approach to capture and predict longitudinal human driving behaviors. It is also leveraged to formulate an efficient ego vehicle model predictive control (MPC) approach to handle random communication delays and packet losses in three different communication topologies: the predecessor following, the predecessor- leader following, and the multi-predecessor following. The proposed approach is compared with several prediction approaches in simulation to demonstrate its effectiveness and find the appropriate communication topology for mixed traffic platoon control. The results show that the predecessor- leader following topology can enhance the benefits of the integrated model predictive control strategy. Specifically, it can lower the average control errors of the following connected and automated vehicles by more than 50% and decrease the control efforts by 10%.