Against the backdrop of the increasing maturity of connected automatic driving technologies and the gradually expanding market share of CAVs, this thesis explores the optimal traffic management strategies to cope with road closures in the context of Connected and Automated Vehicles (CAVs) and Intelligent Transportation Systems (ITS).
A rerouting strategy is designed based on the rerouting behaviour of vehicles when road closure occurs in life, the control parameters include the control of the CAV's automatic rerouting period, rerouting probability, HDV Knowledge of the time of lane closure, as well as their rerouting probability. The aim of this study is to find the optimal combination of these five parameters. Four levels of CAV penetration (20\%, 40\%, 60\%, and 80\%) are considered with the objective of minimizing the total travel time on a mixed CAV and human-driven vehicle (HDV) traffic flow network. The main question is \textbf{What is the optimal rerouting strategy for CAV and HDV mixed traffic when road closure happens?} and in the process of answering this question, the effects of CAV penetration, individual rerouting parameters and different road closure locations are considered and analyzed.
In this thesis, a simulation-based approach is used to model the traffic flow applying both micro and meso scale models. Then, the simulation is conducted for the predefined scenarios, then the sensitivity analysis of each relevant parameter is performed using a one-factor-at-a-time approach to understand the impact of each parameter on the network traffic condition. Finally, Bayesian optimisation is used to find the optimal rerouting strategy within a certain search range and number of times, where the results obtained from the sensitivity analysis are used to determine the parameter search space.
The grid network and the Sioux Falls network are simulated respectively and the relatively optimal rerouting strategies are found for them. The grid network can be regarded as a local area on the network, while the results of Sioux Falls, as a larger network, can provide some basis for city-level traffic management.
The key findings of this thesis include (1) CAV penetration increases bring reductions in TTT and TWT and increase in TTD to the network, overall, the traffic flow movement improves and severe congestion decreases, and network conditions improve significantly during the growth phases of 20\%-40\% and 60\%-80\%; (2) the importance of each rerouting parameter varies for different networks and at different penetration rates, and the results fluctuate significantly between different test values, with no single increasing and decreasing trend; (3) road closures at entrances and exits located at intersections are more critical and require targeted rerouting strategies; the traffic demand distribution has a significant impact on it; (4) Bayesian optimization can find the optimal rerouting strategy in a finite amount of time, where the specific strategy and the improvement effect varies for different networks and levels of demand.