Macroscopic traffic control via connected and automated vehicles in freeway systems

In the last decades traffic congestion in freeway systems has become a major problem, seriously affecting productivity, thus leading to a great social-economic cost. Congestion results in increased travel times for drivers, reduced safety and pollutant emissions. Variable speed limits and ramp metering are some of the most spread techniques for controlling traffic. Recent studies on the future of mobility have highlighted that the automotive world is changing fast, thanks to the newly available technology, and the future of vehicles is expected to move towards connectivity and automation. This introduces the possibility of evaluating new traffic control actions, adopting Connected and Automated Vehicles (CAVs) both as sensors and actuators. In this perspective, conventional traffic models need to be revised in order to capture the impact of smart vehicles on the traffic flow, and specific control actions can be introduced. This research mainly focuses on revising conventional macroscopic traffic models, suitable for traffic control purposes due to their lower computational complexity, in order to capture the presence of smart vehicles. Different modelling approaches are investigated. A first approach consists in modelling each single CAV as it was a moving bottleneck, impacting on the surrounding traffic. A second explored approach considers multi-class mixed human-driven and automated vehicles traffic flows. The great spread of CAVs expected in next years suggests that high penetration rates are likely to appear in traffic systems. In this context, in addition to control each single smart vehicle, the possibility of controlling platooning of CAVs can be envisaged to increase the effectiveness of the traffic control action. Macroscopic models incorporating platoons, belonging to the class of moving bottlenecks models, are then discussed. Once obtained reliable modeling of CAVs moving in the traffic flow, specific control actions are designed. The speed, the number of occupied lanes and the length of platoons can be controlled in cooperative variable speed limits framework, where CAVs act as actuators for the traffic control laws. Adopting macroscopic first-order traffic flow models, although it is a good choice for traffic control applications, introduces the problem of not being able to capture the capacity drop phenomenon. This has also been field of study and a model to describe the capacity drop in first-order models is presented. As last point covered in this thesis, the conventional ramp metering control, already successfully applied worldwide, is revised by applying a sliding mode algorithm. A multi-level hierarchical more complex architecture is also developed to guarantee robustness of the control action in front of disturbances, thanks to the application of the sliding mode algorithms as decentralized controllers, supervised by an higher level model predictive control generating the optimal reference.