Barth Development of Agent - Based On - line Adaptive Signal Control ( ASC ) Framework Using Connected Vehicle ( CV ) Technology

In this study, we developed an adaptive signal control (ASC) framework for connected vehicles (CVs) using agent-based modeling technique. The proposed framework consists of two types of agents: 1) vehicle agents (VAs); and 2) signal controller agents (SCAs) including signal head sub-agent (SH-SA), information processing sub-agent (IP-SA), transition feasibility management sub-agent (TFM-SA) and decision making sub-agent (DM-SA). Within the communication range, each VA communicates with other VAs and SCA and transmits the estimation or prediction of its key statistics, such as position (at the lane level), speed, turning intention and anticipated time-of-arrival (TOA). Then the IP-SA may collect VAs’ statistics and aggregate them into some critical metrics (e.g., queue length, delay, and time utilization rate) at the lane or movement level to support the signal control. With the constraints on phase transition feasibility (e.g., minimum green and movement compatibility), the DM-SA can determine in real-time if the current phase should be extended or switch to another phase. In addition, we proposed a new performance measure, called anticipated green utilization rate (GUR), to evaluate the system performance at traffic signals. Preliminary study in simulation validates the proposed ASC framework using an isolated intersection. The results showed that the ASC algorithms with both queue length optimization and anticipated green utilization rate outperformed the fine-tuned fixed signal timings (with knowledge of hourly traffic demands) in terms of mobility and environment sustainability by the range of 9% 18% and 2% 7%, respectively.