First-best marginal cost toll for a traffic network with stochastic demand
Introduction
It has long been recognized that congestion arises because the marginal social costs of road use diverge from the private costs (Pigou, 1920, Knight, 1924). Therefore, from the theoretical perspective, if a marginal cost toll is allowed to be charged on each link, then this particular first-best toll can drive the traffic flow pattern toward a social optimum (SO) under the routing principle of deterministic user equilibrium (UE). The first-best marginal cost pricing (MCP) has been extensively studies in the literature, in which all links in the network are tolled at a toll level of v dt(v)/dv, where v is the link traffic volume and t is the link travel time function (Yang and Huang, 1998). This simple and elegant formulation for traffic network models provides an important benchmark for road pricing studies. The MCP strategy is also verified to be applicable to the stochastic user equilibrium (SUE) routing principle (Yang, 1999). Maher et al. (2005) formulated a stochastic social optimum (SSO) for the probit SUE model, where the SSO aims to minimize the total expected perceived travel time of networks under the probit SUE condition. In addition to the static form of MCP, a dynamic version was analyzed by Kuwahara (2007), in which the formulation of the dynamic MCP toll is shown to be a direct extension of the static one. Yang and Huang (2004) also extended the analysis of MCP (or system optimum assignment) to the case with multi-class and multi-criteria traffic assignment.
In the abovementioned studies, travel demand is either treated as a deterministic value in the fixed demand case or modeled as a deterministic function of the average travel time (or travel cost) in the elastic demand case. Watling (2002) noted a characteristic feature of all of these models, namely, their deterministic representation of both primary input (e.g., the inter-zonal travel demand matrix and network characteristics) and output (e.g., link flows and travel times) quantity. However, such an assumption may not be plausible or realistic for traffic networks, as they are exposed to various sources of uncertainty both on the demand and supply sides. Several studies have attempted to represent such uncertainties explicitly in network modeling frameworks (stochastic networks). Some of them have focused on representing the stochastic road capacity, which can be affected by major events, such as earthquakes (Bell and Iida, 1997), or relatively minor events, such as adverse weather conditions (Lam et al., 2008). Similarly, many studies have recently begun to include stochastic travel demand (to represent the fluctuation of travel demand from day to day) into the stochastic network model (see, e.g., Clark and Watling, 2005, Shao et al., 2006, Sumalee et al., 2006, Siu and Lo, 2008). To this end, the notion of the stochastic traffic network (SN), as a framework for network modeling with stochastic inputs of travel demand and supply, has emerged. With the SN, equilibrium traffic assignment models can similarly be classified according to the behavioral assumption governing the route choice, i.e., deterministic UE and SO. For brevity, we refer to these models as SN-UE and SN-SO, respectively.
This paper aims to investigate the relationship between the SN-UE and SN-SO models to establish the first-best marginal cost toll scheme for the SN model, as previously done for the deterministic network one. The analysis includes the effect of demand uncertainty on the evaluation of first-best MCP. This problem is practically relevant. For instance, the recent change in the Electronic Road Pricing (ERP) toll adjustment scheme in Singapore involves the consideration of the 85th-percentile traffic condition (speed) to reflect the variability of traffic conditions. The paper also looks at the relationship between the original form of MCP and MCP for the stochastic network (SN-MCP) to identify an additional term caused by stochastic demand. In this paper, for the SN-UE model, we assume that no traveler can reduce his/her expected travel time (risk-neutral traveler) or his/her weighted sum of the mean and the variance of travel time (risk-averse traveler) by unilaterally changing routes. For the SN-SO model, the problem involves distributing the stochastic flows to minimize the expected total travel time in the risk-neutral case and the weighted sum of the mean and the variance of the total travel time in the risk-averse one.
The rest of this paper is organized as follows. Section 2 introduces the notations used in the paper and presents the SN-UE and SN-SO models for the inelastic (fixed) demand case. In Section 3, the relationship between the SN-UE and SN-SO models in the risk-neutral case is discussed, and the SN-MCP model for the general distribution of travel demand is derived. A closed-form SN-MCP formulation for a specific case of the SN with lognormal demand is also presented. The analysis for the risk-averse case is then presented in Section 4 together with the closed-form SN-MCP derivation under the risk-averse model. In Section 5, four numerical experiments are conducted to compare the performance of the network under different MCP principles. Finally, the conclusion and suggestions for future research are given in Section 6. The extension of the general form of the SN-MCP model to the case with elastic demand is provided in Appendix B.
Section snippets
Framework of the stochastic network model
In this section, the mathematical formulations of the SN-UE and SN-SO models for a stochastic network with demand uncertainty are presented.
Analysis of marginal cost pricing
The classical form of MCP was derived under a variety of theoretical frameworks, from the standard case of a homogenous traffic stream moving along a given uniform stretch of road to complex network-based transport model systems. As mentioned, MCP states that road users using congested roads should pay a toll equal to the difference between the marginal social cost and private cost to minimize the total system cost if the demand is fixed, or maximize the social surplus if the demand is elastic.
Model incorporating the risk-based (prone/averse) attitude
In the previous section, travelers are assumed to choose the path that will minimize their expected travel time. This assumption represents the risk-neutral case. The objective function of the SN-SO model is to minimize the expected total travel time, which is also a risk-neutral case for the system manager. Under travel time uncertainty, travelers consider both travel time variability and mean travel time (Bates et al., 2001). In this section, travelers are assumed to consider both the mean
Numerical experiments
This section presents four numerical experiments to illustrate the calculations and implications of the SN-MCP and RSN-MCP toll schemes. The equilibrium mean link flow patterns under different toll schemes are calculated, and the expected total travel times are evaluated and compared.
Conclusions
This paper investigated first-best marginal cost pricing in a stochastic traffic network in which demand uncertainty is explicitly considered. The paper derived the true first-best marginal cost toll for a general network under the assumptions of risk-neutral and risk-averse behavior of travelers in route choice decision. In the risk-neutral case, the marginal cost toll under stochastic demand involves two terms: (i) the mean link flow times the derivative of the mean link travel time with
Acknowledgements
This work was supported by the Research Grants Council of Hong Kong (PolyU 5261/07E) and National Natural Science Foundation of China (Grant No. 70831002). The authors wish to thank the two anonymous referees for their comments and suggestions, which helped to improve the quality of the paper.
References (27)
- et al.
The valuation of reliability for personal travel
Transportation Research Part E
(2001) - et al.
Modelling network travel time reliability under stochastic demand
Transportation Research Part B
(2005) A theory and implications on dynamic marginal cost
Transportation Research Part A
(2007)- et al.
Modeling impacts of adverse weather conditions on a road network with uncertainties in demand and supply
Transportation Research Part B
(2008) - et al.
An efficient approach to solving the road network equilibrium traffic assignment problem
Transportation Research
(1975) - et al.
Stochastic social optimum traffic assignment
Transportation Research Part B
(2005) - et al.
Doubly uncertain transportation network: degradable capacity and stochastic demand
European Journal of Operational Research
(2008) - et al.
Traffic restraint, road pricing and network equilibrium
Transportation Research Part B
(1997) - et al.
Principle of marginal-cost pricing: how does it work in a general network?
Transportation Research Part A
(1998) - et al.
The multi-class, multi-criteria traffic network equilibrium and systems optimum problem
Transportation Research Part B
(2004)
Studies in the Economics of Transportation
Risk-averse user equilibrium traffic assignment: an application of game theory
Transportation Research Part B
Cited by (51)
An empirical study on the stochastic long-term travel demands of a large-scale metro network
2024, International Journal of Transportation Science and TechnologyProportionally fair flow markets for transportation networks
2022, Transportation Research Part B: MethodologicalTravel time reliability-based optimization problem for CAVs dedicated lanes
2022, Transportmetrica A: Transport ScienceMulti-modal urban transit network design considering reliability: multi-objective bi-level optimization
2021, Reliability Engineering and System SafetyCongestion and environmental toll schemes for the morning commute with heterogeneous users and parallel routes
2019, Transportation Research Part B: MethodologicalDesign of a heuristic environment-friendly road pricing scheme for traffic emission control under uncertainty
2019, Journal of Environmental ManagementCitation Excerpt :Meanwhile, the other researchers well addressed the inherent uncertainties in the vehicle EF database and the emission inventory (Pan et al., 2016; Shen et al., 2015; Wang et al., 2008; Yan et al., 2014). Furthermore, from the viewpoint of the transportation network optimization, conventional road pricing models mainly focused on the uncertainties in terms of travel demands and travel time through stochastic and fuzzy programming methods (Lee et al., 2012; Li et al., 2012; Lv et al., 2015; Nikolova and Stier-Moses, 2014; Sumalee and Xu, 2011). Unlike the above studies, traffic emissions are usually characterized by multiple uncertainties due to a variety of influencing factors, which will no doubt present challenges to the existing programming methods.