University researchers develop algorithms and models to improve and evaluate operations on US-29 in Greenville, SC.
Connected Vehicle Supported Adaptive Traffic Control for near-congested Condition in a Mixed Traffic Stream
This study modeled a connected vehicle (CV) based adaptive traffic control system for a 3-mile-long section of US-29 in Greenville, SC. The system design assumed the corridor was equipped with multiple wireless communication devices (Roadside Units) having computation capabilities required to adjust signal timing parameters in real-time based on data collected from connected vehicles transmitting basic safety messages (time, location, speed, direction, etc.) in mixed traffic (CV and non-CVs) on the network.
A machine learning based short-term traffic forecasting model was used to predict the future traffic counts. Platoon movements were estimated and then a Mixed Integer Linear Programming (MILP) model was used to optimize signal intervals at each intersection.
Operational performance improved even for a low CV penetration (5 percent CV) and the benefit increased with increasing CV penetration.
5 percent CV market penetration
- Average speed increased by 5.6 percent
- Average delay decreased by 8 percent
- Average maximum queue length decreased by 66.7 percent
- Average stop delay decreased by 32.4 percent.
100 percent CV penetration
- Average speed increased by 8.1 percent
- Average delay decreased by 13.4 percent
- Average maximum queue length decreased by 70.2 percent
- Average stop delay decreased by 41.4 percent.