Distributed optimization and coordination algorithms for dynamic speed optimization of connected and autonomous vehicles in urban street networks. (October 2018)

• Record Type: Journal Article
• Authors: Tajalli, Mehrdad; Hajbabaie, Ali
• Is Part Of: Transportation research. Volume 95(2018)
• Page Start: 497
• Page End: 515
• Publication Date: 2018-10
• Subjects: Transportation -- Periodicals; Transportation -- Technological innovations -- Periodicals; 388.011
• DOI: 10.1016/j.trc.2018.07.012
• Languages: English
• ISSNs: 0968-090X

Abstract

Highlights: Introduced a dynamic speed optimization method in urban-street networks with autonomous vehicles. Developed distributed optimization and coordination algorithms. The methodology finds near-optimal solutions to dynamic speed optimization in real-time. The observed optimality gap was at most 2.7%. Speed optimization reduced the travel time by up to 14.8% and speed variation by 9.7–13.4%. Abstract: Dynamic speed harmonization has shown great potential to smoothen the flow of traffic and reduce travel time in urban street networks. The existing methods, while providing great insights, are neither scalable nor real-time. This paper develops Distributed Optimization and Coordination Algorithms (DOCA) for dynamic speed optimization of connected and autonomous vehicles in urban street networks to address this gap. DOCA decomposes the nonlinear network-level speed optimization problem into several sub-network-level nonlinear problems thus, it significantly reduces the problem complexity and ensures scalability and real-time runtime constraints. DOCA creates effective coordination in decision making between each two sub-network-level nonlinear problems to push solutions towards optimality and guarantee attaining near-optimal solutions. DOCA is incorporated into a model predictive control approach to allow for additional consensus between sub-network-level problems and reduce the computational complexity further. We applied the proposed solution technique to a real-world network in downtown Springfield, Illinois and observed that it was scalable and real-time while finding solutions that were at most 2.7% different from the optimal solution of the problem. We found significant improvements in network operations and considerable reductions in speed variance as a result of dynamic speed harmonization.