A novel framework for real-time adaptive signal control using connected vehicles

Intelligent transportation systems (ITS) work by collections of data in real time. Aver-age speed, travel time and delay at intersections are some of the most important measures, often used for monitoring the performance of transportation systems, and useful for system management and planning. In urban transportation planning, inter-sections are usually considered critical points, acting as bottlenecks and clog points for urban traffic. Thus, detecting the travel time at intersections in different turning di-rections is an activity useful to improve the urban transport efficiency. Smartphones represent a low-cost technology, with which is possible to obtain information about traffic state. However, smartphone GPS data suffer for low precision, mainly in urban areas. In this work, we propose a novel framework for real-time adaptive signal control us-ing connected vehicles (CV). This framework is made of two new methods: the first one is for lane identification and flow estimation, in which we aim to determine the traffic flows on lanes near the intersections controlled by traffic lights, starting from GPS data acquired by smartphones; the second one is for optimal real-time traffic signal settings. In the first method, we present a fuzzy set-based method for identification of vehicles position within road lanes near intersections using GPS data coming from smartphones. We have introduced the fuzzy sets to consider uncertainty embedded in GPS data when trying to identify the position of vehicles within the lanes. Moreo-ver, we introduced a Genetic Algorithm to calibrate the fuzzy parameters and to ob-tain a novel supervised clustering technique. In more details, for each lane identified by the proposed method, we set up a flow estimation method to optimize signal timings. It is based on the evaluation of the length of the queues and the speed. Finally, the proposed method has been validated through the comparison with the fundamental diagram. Signal timings were optimized using the Webster algorithm. As for the case study, we have studied a signalized intersection in the city of Bari (Italy), considering three main time periods: 8 a.m. – 9 a.m., 12 a.m. - 1 p.m. and 5 p.m-6 p.m. We acquired data related to location, speed, travel times and trajectories of a vehicle using a smartphone application. Smartphone devices have the ad-vantages of mobile sensors: low investment costs, high penetration, and high accu-racy achieved by GPS receivers. In addition, GPS-enabled smartphones can provide accurately not only the position but also speed and travel direction. First results reveal the effectiveness of the proposed method about the lane identifi-cation in comparison with the outcomes of two well-known clustering techniques (Fuzzy C-means, K-means). The curves calculated by the Greenshields method are comparable with the funda-mental diagrams. Results regarding the signal optimization show that the improve-ments obtained by our method are remarkable: in some cases, we have achieved improvements around 50% for the delay times and in the reduction of average length of queues.