ADAPTIVE CONTROL STRATEGY FOR TRAFFIC PLATOON COORDINATION EMPOWERED BY CONNECTED AND AUTOMATED VEHICLES

Abstract

Traffic congestions caused by increasing population, car ownership, and corresponding traffic demands have been a serious issue that not only hazards the safety and comfort of the commuters, but also increases the economic losses and environmental damages. Platoon control enabled by connected and autonomous vehicles (CAV) is regarded as a powerful solution to alleviate the traffic pressure and congestions caused by road bottlenecks. However, it will take a long time before the CAV can share a high market penetration rate to achieve pure CAV platooning. Therefore, both CAVs and the human driven vehicles (HDV) without Vehicle-to-Vehicle (V2V) communication mechanism will coexist and form mixed platoons in the future for a long period of time. It places urgent demands on reliable and robust strategies for this kind of mixed platoons to achieve steady and smooth platoon control. In this study, a specific mixed platoon composition with a leading CAV, a following CAV, and several sandwiched HDVs between them is adopted to research the mixed platoon characteristics. A spring-mass-damper-clutch physical system inspired linear car following model with delay is used to account for the delayed reaction and the varying driving modes of human drivers. Since the HDVs are not equipped with V2V communication devices, the last following CAV is regarded as degraded CAV (DCAV). Therefore, an Adaptive Numerical algorithm for Subspace system Identification (AN4SID) method is proposed to identify the model of direct preceding HDV in front of the last following DCAV. The model could be used to predict the future states of the preceding HDV given the estimated delay through vector fitting and the planned future trajectory of the leading CAV. Then a hybrid control strategy combining the Model Predictive Control (MPC) and the optimal feedback control methods is applied to the following DCAV. The predicted HDV states will be used to calculate the reference trajectory of the following DCAV for the MPC process. Then the MPC control signal will be used as the feedforward input together with the feedback control input to eliminate the deviation caused by prediction errors.By conducting simulations based numerical experiments on MATLAB platform and the real vehicle trajectory data from the Next Generation Simulation (NGSIM) open dataset, the proposed AN4SID method shows reliable prediction performance compared with the Recursive Least Square (RLS) method and the Iterative Extended Kalman Filter (IEKF) method. Then, further experiments judged that our proposed hybrid control strategy overperformance both the Linear Quadratic Regulator (LQR) and the regular MPC algorithm, in terms of the safety, comfort of passengers, and fuel efficiency.