01628871

Component

Weaving sections constitute critical roadway bottlenecks along freeways and highways. The intense lane-changing behavior along these sections significantly reduces the roadway capacity. With the development of wireless communication technologies, connected vehicles (CVs) may be leveraged to develop advanced traffic management systems to increase the capacity of weaving sections. In this paper, the authors develop a ramp-metering control algorithm based on reinforcement learning techniques to increase the weaving section capacity. The control system monitors weaving section arrival rates from connected vehicles and searches for the optimal ramp-metering rate using a novel Q-learning algorithm. Simulation of a single weaving section showed that the system increases the weaving section capacity by approximately 3.5% and significantly mitigates congestion on the mainline freeway. The second example of two weaving sections shows even larger savings. Specifically, the maximum discharge flow rates through the two weaving sections increases by 12% and 20%, respectively. The system also reduces the network-wide fuel consumption and greenhouse gas emission levels by approximately 10%.

This paper was sponsored by TRB committee ABJ70 Standing Committee on Artificial Intelligence and Advanced Computing Applications.

01618707

17-03689

English

20p

2017

Transportation Research Board 96th Annual Meeting

Digital/other

Figures; References; Tables

Algorithms; Control systems; Highway capacity; Machine learning; Mobile communication systems; Ramp metering; Weaving sections

Highways; Operations and Traffic Management

Transportation Research Board Annual Meeting 2017 Paper #17-03689

TRIS, TRB, ATRI

Dec 8 2016 11:24AM