Analysis of current approaches in optimization of transport network management

One of the key objectives of the TANGENT project is to propose advanced techniques for modelling and simulation, optimization and control that can capture the dynamics of traffic and demand and adapt to evolving complex multimodal transport settings. In this complex environment, the optimization of traffic management and the multimodal transport network plays an important role since it allows users to have better door-to-door mobility and transport authorities to improve the management of the transport network.
This deliverable aims to analyze scientific literature related to current approaches to transport network management optimization within a multi-actor setting. Concretely, we focus our literature review on the three following problems because they play a relevant role in the management of mobility at a network level, where different actors and means of transport must be coordinated:
• Signal Vehicle Couple Control with CAVs. It aims to improve the traffic control performance by leveraging the exchange of information in real-time between signals and vehicles (connected vehicles and CAVs), and the simultaneous optimization of signals timing/phases and CAVs trajectories and/or routes, to enhance the performance of the whole traffic network.
• Synchronization of shared and on-demand mobility with transit modes. It is an area of vital importance because it can contribute to public transport's future position as the backbone of mobility in urban areas. Concretely, it can provide a solution to addressing the first/last mile problem of transit modes, especially in areas not densely populated and with infrequent public transport services since these modes can act as feeders or collectors of public transit.
• Dynamic Congestion Pricing. It is a variant of congestion pricing where prices vary dynamically in real-time as a function of current traffic conditions. This approach is opposed to flat pricing, which stays constant over time, and scheduled pricing, where tolls vary by time of day, day of the week or season following a predetermined schedule.
In the review of the literature, we analysed the optimization models and techniques commonly used in the scientific literature reviewed for the three categories of problems mentioned above. Here it is important to mention that the literature review focused only on papers that address those problems from a pure optimization perspective. In this analysis of the literature, we considered aspects such as the application scope, the optimization models (decision variable, constraints, objective function and modelling approach) and methods used, the experimental benchmark used and the comparison studies performed. From this literature review, the main research gaps identified for each problem were the following:
• Signal Vehicle Couple Control with CAVs
o Most of the literature considered a single objective for the optimization purpose but inherently the SVCC problem is multi-objective in nature.
o Most of the studies published so far are conducted under a 100% CAV scenario or a mixed traffic scenario with a high penetration rate of CAVs.
o Only a few studies focus on the optimization of SVCC at the corridor or network level.
• Synchronization of shared and on-demand mobility with transit modes
o Only a few papers addressed the synchronization of shared- and on-demand mobility with public transit in rural scenarios.
o Lack of studies considering shared or on-demand services based on CAVs for their synchronization with public transport.
o Most of the studies focus on synthetic data which can lead to biased conclusions.

• Dynamic Congestion Pricing
o Most of the current DCP schemes are based on single-objective models.
o Only a few papers address large-scale applications of DCP schemes because of their computational complexity.
Based on the literature review previously mentioned and on the priorities defined by the different case studies in Deliverable D1.1, in this document, we also provide a description and a justification of the particular optimization problems in transport network management in which we will focus on the upcoming tasks of TANGENT. Concretely, the chosen optimization problems are:
• Coupled traffic signal and route planning optimization for CAVs. Concretely, the objective is the coupled optimization (considering multiple objectives) of traffic signal control and CAVs schedules and routes at corridor level (at least) and under a mixed traffic scenario.
• Optimization of integration DRT systems with public transit modes. More specifically, the aim is the joint optimisation of the capacity (frequency and size of allocated vehicles) of public transport lines and the prioritisation of public transport at signalised intersections based on dynamic transit assignment models.
• Synchronization of public transport and Traffic control. In this case, the objective is the joint optimisation of the capacity (frequency and size of allocated vehicles) of public transport lines and the prioritisation of public transport at signalised intersections based on dynamic transit assignment models.
• Optimization of Dynamic Congestion Pricing schemes. Concretely, here we aim at the optimization of DCP schemes using a multi-objective and within-day approach making use of model-based optimization and parallel computing.
In this document, we also review literature related to negotiation and arbitration models for transport network management. Concretely, we focus on integrated decision-making, given that designing and operating urban transportation systems is a complex process that must consider various factors (e.g. economic, environmental, and socio-political), and that entails the involvement of various stakeholders with their own objectives and priorities. However, involving different stakeholders in the decision-making process has proven benefits for the multiple components that constitute the urban transport network. For this reason, in this deliverable, we discuss different consensus definitions, distinguishing between a “hard” and “soft” interpretation of consensus, and we examine ways to measure it. We also review traditional approaches to selecting the optimal solution when the stakeholders have different preferences focusing on Cost-Benefit Analysis methods, Multi-criteria Decision-making techniques and preference-based Evolutionary Algorithms. After that, we also discuss the state-of-the-art in decision-making approaches based on the principles of Agent-Based Modelling, focusing on Agent-Based Negotiations and Agent-Based Social Dynamics. Finally, considering all the review literature on this topic, we provide practical guidelines that delineate the development of a consensus-reaching mechanism, discussing its objective, scope, domain of application, stakeholders, data needs and modelling approach, and how they will be applied in the context of the TANGENT project.
Finally, we also overview software tools that can be used for the optimization of transport network management, which can be grouped into two categories, namely, machine learning-oriented (SMAC, Keras, TensorFlow, TensorFlow Model Optimization, Scikit-learn), and optimization-oriented (DEAP, PyMoo7 and Nevergrad)