THE CELL TRANSMISSION MODEL FOR URBAN TRAFFIC : A ROBUST OPTIMIZATION ALGORITHM FOR SIGNAL SETTING PROBLEM

Abstract

The main goal of this research is to define a robust method to solve combined signal setting problem. Signalized junctions represent critical points of an urban transportation network, and the efficiency of their traffic signal setting in uences overall network performance. Since road congestion usually takes place at or close to junction areas, an improvement in signal settings contributes to improve travel times, driver comfort, fuel consumption e ciency, pollution and safety. In a tra c network, the signal control strategy a ects the travel time on the roads and in uences the drivers' route choice behavior. Usual traffic signal optimization methods seek either to maximize the green bandwidth or to minimize a general objective function that typically includes delays, number of stops, fuel consumptions and some external costs like pollutant emissions. A major objective of Traffic Signal Synchronization at intersection is to clear maximum tra c throughput in a given time with least number of accidents, at maximum safe speed and with minimum delay. It has been widely accepted that improving traffic flow has been one of the strategies to reduce vehicle emissions and fuel consumption. In urban areas, frequent stop-and-go driving and excessive speed variations contribute to higher fuel consumption and emissions. During the implementation of the research project presented in this PhD Thesis, three main areas of research have been explored: urban tra c modeling, tra c control, network clustering. This thesis presents a new framework for tra c modeling, based on Cell Transmission Model. The Cell Transmission Model for Urban Tra c (CTM-UT) proposes new methods to model complex tra c dynamics on arterial and node. Its two most important features consist in the representation of (i) the tra c ow interactions between neighboring movements when entering in the channelized lanes upstream intersection and (ii) the conflict among crossing flows at intersection with the capacity determination of minor flows. The experiments (presented in [Adacher et al., (2014)] , [Adacher and Tiriolo (2015b)] and [Adacher and Tiriolo (2015b)]) demonstrate that the CTM-UT is suitable to obtain good accuracy (between the 2% and 4%) compared to microscopic simulation tools (i.e., VISSIM and SUMO). At the same time, the CTM-UT framework is suitable to take into account the urban complexity keeping the bene ts of macroscopic model (i.e., low computational time and not too many information to manage). To improve the calibration process of macroscopic models, which require a great e ort in terms of time and resources, a new method to calibrate the congestion wave speed via optimization is proposed. In order to minimize total tra c delay with signal synchronization, [Adacher and Tiriolo (2016c)] describe how the Surrogate Method (SM), based on robust gradient approach, has been improved. The new version of the optimization model is suitable to obtain solution with more e ciency (at least the 4%) and e cacy (at least the 10%) compared to the classic SM. Lastly, new "hybrid" decompositions of network nodes are proposed and the classi cation of nodes is integrated with a clustering algorithm based on topological characteristics of the network. Given the network partitions (subnetworks), the SM is applied to solve the traffic signal synchronization problem. Compared to considering the total network, the optimization integrated with the new "hybrid" decomposition approaches can reduce the computational time with a slight loss of accuracy. For example, on a small network, an improved e ciency of 78% and a e cacy loss of 0:6% have been obtained (see [Adacher and Tiriolo (2016a)] and [Adacher and Tiriolo (2016b)]). While, on a large network, an improved e ciency of 53% and a e cacy loss of 0:09% have been obtained (see [Adacher and Tiriolo (2016c)]). The results presented in this thesis demonstrate that the new version of optimization model, applied with proposed decomposition algorithm, are suitable to minimize, through the tra c signal synchronization, the total network delay in real time and for large-scale networks. The tra c model proposed is applied to estimate, with accuracy and e ciency, the total delay taking into account of complex urban dynamics. The knowledge of travel time variations is fundamental in the evaluation of Dynamic Tra c Assignment (DTA) strategies and in the travelers' choices. Since that the CTM-UT is suitable to predict the travel time for tra c demands detailed at urban level (link, lane, turning movement), an interesting future research will be focused to the application of the DTA combined to the new proposed optimization algorithm. The integration of the assignment model into the de ned control strategy might lead to increase the computational time but this problem can be reduced with the presented decomposition approaches.