Design, optimization and simulation of trad­able mobility credits

Road traffic congestion is a critical problem affecting urban mobility worldwide and its severity continues to increase, causing significant costs at the individual, environmental, economic, and societal levels. While a significant agenda has been put forward on the transport supply side, mostly driven by vehicle technology (automation and electrification), demand shifts are often considered a hard-to-reach but effective means to reduce the social and environmental costs associated with transport. Demand management has thus become an increasingly important focus of the policy agenda in many metropolitan areas. Congestion pricing as a demand management instrument has been widely investigated in both theory and practice motivated by its potential gains in social welfare. Nevertheless, congestion pricing often receives political and social resistance as it is perceived as a tax and in some contexts, can be vertically inequitable. An alternative market-based solution called a tradable credit scheme (TCS) has been receiving attention in recent years. In a typical TCS system, a regulator predetermines a total quota of credits available for the area and period of interest and distributes these credits to all potential travelers. The credits can be bought and sold in a free market at a price determined by credit demand and supply. Consequently, a tradable credit scheme has mainly three potential advantages over congestion pricing (without revenue redistribution): (i) TCS is revenue neutral as there is no monetary transfer to or from the regulator; (ii) TCS can be more equitable than congestion pricing since the inconvenience caused by the limited use of vehicles is compensated by selling extra credits; (iii) TCS has been shown to yield efficiency gains under uncertainty when congestion is significant. The first two features of TCS could help address the long-standing issue of public opposition to congestion pricing. This PhD study extends the growing body of literature in TCS with new area-based TCS designs, the flexible modeling and assessment of TCS via agent-based simulations, and the development of TCS optimization frameworks using machine learning techniques. This thesis is divided into three parts: i) Part I presents two studies on the formulation and application of Bayesian Optimization (BO) to the second-best design of tariff schemes for congestion pricing (and used in the design of tariff schemes for tradable mobility credits in Part II), ii) Part II includes two studies proposing different trading mechanisms (peer-to-regulator and peer-to-peer) for trip- and area-based TCS for the management of urban networks, and iii) Part III proposes a detailed and flexible simulation framework for assessing the impact of different TCS designs under realistic scenarios by extending a state-of-the-art activity-driven agent-based simulation platform (SimMobility). Part I deals with the development of two BO frameworks for congestion pricing optimization problems with different perspectives on utilizing problem-specific information for efficiency improvement. In the first study, we propose a BO formulation with problem-specific dropout strategies which can learn the relationship between the tariffs (decision variables) and social welfare (objective function) within a few iterations even under high-dimensional tariff structures. In the second study, we further develop a contextual BO framework where the BO scheme is embedded within the day-to-day dynamic model by using temporal contextual information. We numerically demonstrate that the framework utilizes a significantly smaller number of simulation evaluations (ten-fold reduction) than the standard BO approach. This framework can also incorporate context-specific demand and supply information which can be of value to policymakers when evaluating optimal tariff design schemes under a wide range of scenarios in a computationally tractable manner. We further show that distance-based tariff schemes yield significant welfare gains relative to area-based schemes and highlight that the design of the distance-based tariff scheme can significantly affect distributional impacts: a suitably designed two-part tariff structure can partially offset the relatively large welfare losses of travelers with longer commute distances while maintaining overall welfare. Part II focuses on the design and properties of TCS when applied to trip-based Macroscopic Fundamental Diagram (MFD) models considering the dynamics of the credit price. We propose an area-based TCS with time- and distance-based credit tariffs and incorporate this TCS into a day-to-day modeling framework with heterogeneous travelers. Analytically, we present conditions for the existence of the market and network equilibrium and the uniqueness of the credit price. Numerical results validate the analytical properties (including convergence and credit price uniqueness and its inverse proportionality with the credit allocation), demonstrate that the proposed TCS yields identical social welfare as congestion pricing while maintaining revenue neutrality, and show the superiority of a trip-based TCS to a trip-agnostic area-based TCS. Here, all travelers are assumed to trade directly with the regulator (peer-to-regulator, P2R), where, in the short-term, budget neutrality of credits cannot be guaranteed. Instead, these issues can be partially addressed by a peer-to-peer (P2P) trading market design where the transactions of credits happen between travelers. Although most existing TCS research adopts the assumption of P2P trading, the underlying mechanism that achieves market clearing (in terms of matching sellers and buyers and pricing of credits) is typically not elaborated. Thus, in Part II's second study, we investigate two different P2P trading paradigms that define the rules of matching selling and buying orders, market price adjustment, and individual bidding format. Numerical results show that all proposed P2P trading paradigms lead to a near identical equilibrium in terms of social welfare gains, departure flows, and credit price as that obtained from P2R schemes, while the P2P trading mechanisms are able to ensure the budget neutrality of credits as well as revenue neutrality of the regulator during the day-to-day process. Most TCS studies to date use simple demand and supply models for assessment. A TCS however may affect several mobility related dimensions of decision making at the individual level and interactions of users, influencing demand patterns, traffic flows, and network performance. Agent-based simulation is a common technique to capture these complex interactions and conduct analysis and evaluation of different TCSs. In Part III, we propose a flexible framework with a modular and extensible implementation in the state-of-the-art urban simulator SimMobility for the detailed simulation of the operation of a TCS system. Demand is modeled through an activity-based model and a within-day departure and route choice model sensitive to individual TCS, account budgets and heterogeneous preferences. The transportation supply is now represented by a mesoscopic network model and is extended with a TCS controller for handling all credit transactions within the simulation. This proposed framework allows for the simulation of a variety of TCS design schemes and is tested in a prototypical urban setting where theoretical TCS properties are then assessed. Another contribution of this work is that the developed scalable, operational, flexible, and open-source simulation platform is delivered as part of the PhD project. In summary, this PhD study contributes to the body of literature in mobility demand management, namely in tradable credit schemes, covering the topics of machine-learning based optimization, market design for TCS, and the design and assessment using more realistic demand and supply models. This thesis provides promising simulation-based optimization approaches for the optimal design of tariffs, enabling the efficient design of demand management instruments. The findings of this thesis also bring insights into the properties of the area-based TCS as well as key modeling and implementation frameworks for the design of future TCS including trading behaviors, credit allocation, expiration, trading patterns, price adjustment, and complex behavioral changes.Doctor of Philosoph