Control of connected and autonomous vehicles

Abstract

It is hard to believe that less than 150 years ago, Malta was an island without a single vehicle. Today, the Maltese population is highly dependant on vehicles for transportation. This is resulting in traffic congestion on a daily basis in multiple localities. Traffic provides a trigger to several other problems, one of which is poor air quality resulting in increased health risks. Therefore, it is in everyone’s interest that traffic should be reduced drastically and health hazards possibly eliminated completely. The development of Connected and Autonomous Vehicles (CAVs) is a significant breakthrough that could lead towards mitigating or even eliminating traffic build up. The aim of this study is to simulate and control the flow of CAVs through a local signalized urban junction. The central intersection considered in this study, is the Rue D’Argens and Sliema road junction. However, other junctions that are in the vicinity of this intersection are also included in the test network model. To draw a realistic comparison between the implemented CAV algorithms and the current state of the area, traffic flow readings are obtained from the selected roads. These traffic readings are then used to calibrate the simulated network according to actual flows. Since this study is a continuation on a previous project, the best performing traffic management system out of the four algorithms implemented, was adapted for this project. This algorithm collects the required data from CAVs approaching the central junction and groups them into platoons. The traffic light timings are also controlled so that platoons can pass through the junction uninterrupted. In this dissertation, the behaviour of CAVs is also studied and a new car-following model is implemented to the simulated network model. By comparing the results obtained from the calibrated junction area and the platooning algorithm, statistical improvement is shown for all CAV market penetration rates greater than 40%. Improved performance is also observed during weekday tests with the addition of the new car-following model to the platooning algorithm.