Vehicle-Traffic Control with Limited-Capacity Connected/Automated Vehicles

This project studies methods to control both vehicle and traffic under limited penetration of low-level connected and autonomous vehicles (LCAVs). The investigation includes three major parts: (1) the Eco-Driving algorithm for a single CAV with low-level automation; (2) the vehicle in the loop (VIL) simulation platform; and (3) the integrated vehicle/traffic control algorithm tested under the VIL. A hybrid deep Q-learning and policy gradient (HDQPG) based Eco-Driving algorithm is first developed for a single CAV driving along signalized corridors with low-level automation to learn and control both the longitudinal operations and the lane-changing decisions of the LCAV. Second, a vehicle in the loop (VIL) platform is developed to reduce the costs of testing algorithms in real-world. Third, a dynamic Highway Capacity Manual (HCM) method is designed to control signals under the environment of LCAVs. The method first estimates the CAV penetration and calculates the total traffic volume based on the estimated CAV penetration and the CAV volume. Given the volume data, the method calculates the signal phases/timings based on the HCM method, which can be used for the next forward time horizon. The authors combine this dynamic HCM method with the Eco-Driving algorithm mentioned as a vehicle-traffic integrated control method, and test it on the VIL platform. The results show that the integrated method could reduce fuel consumption of the controlled vehicle and reduce delays of all vehicles of the corridor.


  • English

Media Info

  • Media Type: Digital/other
  • Features: Appendices; Figures; Photos; References; Tables;
  • Pagination: 64p

Subject/Index Terms

Filing Info

  • Accession Number: 01767121
  • Record Type: Publication
  • Created Date: Mar 11 2021 9:29AM