Dynamic programming-based multi-vehicle longitudinal trajectory optimization with simplified car following models

Yuguang Wei, Cafer Avcı, Jiangtao Liu, Baloka Belezamo, Nizamettin Aydın, Pengfei(Taylor) Li, Xuesong Zhou

Research output: Contribution to journalArticle

30 Scopus citations

Abstract

Jointly optimizing multi-vehicle trajectories is a critical task in the next-generation transportation system with autonomous and connected vehicles. Based on a space-time lattice, we present a set of integer programming and dynamic programming models for scheduling longitudinal trajectories, where the goal is to consider both system-wide safety and throughput requirements under supports of various communication technologies. Newell's simplified linear car following model is used to characterize interactions and collision avoidance between vehicles, and a control variable of time-dependent platoon-level reaction time is introduced in this study to reflect various degrees of vehicle-to-vehicle or vehicle-to-infrastructure communication connectivity. By adjusting the lead vehicle's speed and platoon-level reaction time at each time step, the proposed optimization models could effectively control the complete set of trajectories in a platoon, along traffic backward propagation waves. This parsimonious multi-vehicle state representation sheds new lights on forming tight and adaptive vehicle platoons at a capacity bottleneck. We examine the principle of optimality conditions and resulting computational complexity under different coupling conditions.

Original languageEnglish (US)
Pages (from-to)102-129
Number of pages28
JournalTransportation Research Part B: Methodological
Volume106
DOIs
StatePublished - Dec 2017

Keywords

  • Autonomous vehicle
  • Car-following model
  • Traffic flow management
  • Vehicle trajectory optimization

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Transportation

Fingerprint Dive into the research topics of 'Dynamic programming-based multi-vehicle longitudinal trajectory optimization with simplified car following models'. Together they form a unique fingerprint.

  • Cite this