Platoon-centered control for eco-driving at signalized intersection built upon hybrid MPC system, online learning and distributed optimization part II: Theoretical analysis
Platoon-centered control for eco-driving at signalized intersection built upon hybrid MPC system, online learning and distributed optimization part II: Theoretical analysis
复制标题
基于混合MPC系统、在线学习和分布式优化的信号交叉口以队列为中心的环保驾驶控制第二部分：理论分析
DOI：
10.1016/j.trb.2023.03.008
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Du, Lili
中科院分区：
文献类型：
--
作者：
Zhang, Hanyu;Du, Lili
受互联和自动驾驶汽车（CAV）技术的启发，广泛的研究开发了开环车辆级轨迹规划或速度咨询，以促进交通路口的生态驾驶。但是，很少有研究的排级闭环轨迹控制，可以更好地维持流交通的畅通性和效率。基于这一研究空白，本研究开发了一种系统最优的以车队为中心的生态驾驶控制（PCC-eDriving），它可以引导由互联和自动驾驶车辆（CAV）和人类驾驶车辆（HDV）混合的车队顺利接近，根据需要分开，然后依次通过信号交叉口，同时减少甚至避免急剧减速和红色怠速。为了防止文章冗长，将其分为第一部分和第二部分。具体而言，本研究的第一部分将PCC-eDriving建模为混合模型预测控制（MPC）系统。它涉及三个MPC控制器的队列轨迹控制和混合整数非线性规划（MINLP）的最优分割决策。每个MPC控制器都集成了鲁棒车辆动力学和在线自适应曲线学习算法，以考虑控制和车辆驾驶的不确定性。为了有效地分布式求解MPC控制器的大规模优化问题，提出了一种基于有效集的最优条件分解算法（AS-OCD）。基于现场数据和模拟数据的数值实验表明，PCC-eDriving能够显著提高城市信号交叉口的交通顺畅性和效率，同时降低能耗和排放。第二部分将分析和证明混合MPC系统的时序可行性和输入-状态稳定性，以及AS-OCD求解方法的收敛性，以理论上维持混合MPC系统的性能。
Inspired by connected and autonomous vehicle (CAV) technologies, extensive studies have developed open-loop vehicle-level trajectory planning or speed advisory to promote eco-driving at traffic intersections. But few studies work on platoon-level closed-loop trajectory control, which can better sustain stream traffic smoothness and efficiency. Motivated by this research gap, this study developed a system optimal platoon-centered control for eco-driving (PCC-eDriving), which can guide a platoon mixed with connected and autonomous vehicles (CAVs) and human-driven vehicles (HDVs) to smoothly approach, split as needed, and then sequentially pass signalized intersections, while reducing or even avoiding sharp deceleration and red idling. The effort is separated to Part I and Part II to prevent a lengthy article. Specifically, Part I of this study modeled the PCC-eDriving as a hybrid Model Predictive Control (MPC) system. It involves three MPC controllers for platoon trajectory control and a mixed-integer nonlinear program (MINLP) for optimal splitting decisions. Each MPC controller is integrated with robust vehicle dynamics and an online adaptive curve learning algorithm to factor control and vehicle driving uncertainties. An active-set-based optimal condition decomposition algorithm (AS-OCD) was developed to efficiently solve the MPC controllers' large-scale optimizers in a distributed manner. The numerical experiments built upon the field and simulated data indicated that the PCC-eDriving could significantly improve traffic smoothness and efficiency while reducing energy consumption and emission at urban signalized intersections. Part II will analyze and prove the sequential feasibility and the Input-to-State stability of the hybrid MPC system, as well as the convergence of the AS-OCD solution approach to theoretically sustain the performance of the hybrid MPC system.