Achieving System Optimal Routing and Understanding Implication to Travel Demand Forecasting [ITM # 82]

Transcription

1 Paper Author (s) Xianbiao Hu (corresponding), University of Arizona Yi-Chang Chiu, University of Arizona Jeffrey A. Shelton, Texas A&M Transportation Institute Paper Title & Number Achieving System Optimal Routing and Understanding Implication to Travel Demand Forecasting [ITM # 82] Abstract This paper documents the research effort in developing a Behaviorally Induced System Optimal model to improve the system performance towards System Optimal. Most existing real-time traveler information systems use real time travel time when searching for the shortest path without explicitly considering the marginal cost to the system when assigning multiple travelers to the system. The proposed approach includes the development of a marginal cost calculation algorithm, a time-dependent shortest path search algorithm, and schedule delay as well as optimal path finding modules. Both analytical derivation and numerical analysis have been conducted on CAMPO network in Austin TX as part of the mobility investment priorities analysis report prepared for Texas Transportation Commission and 83rd Texas Legislature, in which report it was built off the core scenario (SC2) as supplemental analysis (SC2d). The case study results show the benefit of the proposed methodology in producing significant traffic congestion alleviation, reduced travel time and monetary saving. Statement of Financial Interest The co-authors including Dr. Xianbiao Hu at Metropia Inc., Professor Yi-Chang Chiu at the University of Arizona (UA), and Mr. Jeff Shelton at the Texas A&M Transportation Institute (TTI) are well-known researchers who have devoted their careers in the advancement of network modeling research and applications. The goal of submitting this research brief is to shed light in understanding how the emerging Advanced Traveler Information System (ATIS) services may enable the system optimal routing and what the resulting implications are from a travel demand forecasting standpoint. All the co-authors do not have any financial interest in the acceptance of this research brief as the authors don t promote or sell any product or services that would be affected by the acceptance of this research brief. Statement of Innovation Most existing real-time traveler information systems use real time travel time when searching for the shortest path without explicitly considering the marginal cost to the system when assigning multiple travelers to the system. The basic design concept is to present generic information to travelers, leaving travelers to react to the information their own way. This passive way of managing traffic by providing generic traffic information makes it difficult to predict outcome and may even incur adverse effects,

2 such as overreaction (a.k.a. the herding effect). The innovation of this research is in how to provide users with a route that aims to minimize the marginal system impact that results from this routing. In other words, when one makes a trip, the true marginal cost (MC) would include not only a single traveler s experienced travel time in the network, but also the delays this trip imposes upon other vehicles in the vicinity or departing afterward. In this paper, a behaviorally-induced, system optimal (SO) model is presented; aiming to further improve the system level traffic condition towards SO through incremental routing. The proposed approach includes the development of a marginal cost calculation algorithm, a time-dependent shortest path search algorithm, and schedule delay as well as optimal path finding modules. In a real-world application, such routes can be provide to users through some incentive mechanisms so that a certain percentage of drives who are willing to deviate from their original planned trip can be rewarded by taking the SO route. Further, from a travel demand forecasting standpoint, it is important to start recognizing the implication of such emerging services to the network models which are the integral component of either trip-based or activity-based framework. Such a network-level impact and modeling implications are fully depicted through a real-world case study for Austin, TX as part of the mobility investment priorities (MIP) analysis prepared for Texas Transportation Commission and 83rd Texas Legislature, in which report it was built off the core scenario (SC2) as supplemental analysis (SC2d). The case study results show the benefit of the proposed methodology in producing significant traffic congestion alleviation, reduced travel time and monetary savings.

3 Achieving System Optimal Routing and Understanding Implication to Travel Demand Forecasting Xianbiao Hu, PhD, Metropia Inc. Yi-Chang Chiu, PhD, University of Arizona Jeff Shelton, Texas A&M Transportation Institute 1 INTRODUCTION Real-Time Traveler Information Systems or Advanced Traveler Information Systems (ATIS) provide pre-trip and/or en route information allowing travelers to quickly assess and react to unfolding traffic conditions. The basic design concept is to present generic information to travelers, leaving travelers to react to the information their own way. This passive way of managing traffic by providing generic traffic information makes it difficult to predict outcome and may even incur adverse effect, such as overreaction (a.k.a. the herding effect). For those ATIS that come with path finding functionality, the goal is to provide users with the path with lowest instantaneous travel time (cost) without explicitly considering the marginal impact of this routing action or the actions of many others following similar requests. The research interest discussed in this paper is how to provide users with a route that aims to minimize the marginal system impact that results from this routing. In other words, when one makes a trip, the true marginal cost (MC) would include not only the one traveler s experienced travel time in the network, but also the impact this trip imposes upon other vehicles in the vicinity or departing afterward. In this paper, a behaviorally-induced, system optimal (SO) model is presented aiming to further improve the system level, traffic condition towards SO through incremental routing. Both analytical derivation and numerical analysis have been conducted on the Capital Area Metropolitan Planning Organization (CAMPO) network in Austin, TX, as part of the mobility investment priorities analysis report prepared for Texas Transportation Commission and 83 rd Texas Legislature, in which report it was built off the core scenario (SC2) as supplemental analysis

4 (SC2d). The outcome of this study shows that our proposed modelling framework is promising for improving network traffic conditions towards SO, resulting in a vast amount of economic savings. 2 METHODOLOGY In exploring the overall concept, this research chooses to apply simple marginal calculation combined with the time-dependent shortest path and path finding method as an integrated procedure. The MC of concern includes both route and departure time dimensions. From the departure time standpoint, this research considers that different trip purposes may have different departure or arrival time flexibility constraints. The preferred decision for work trips for example, may lean toward either leaving earlier or taking a less congested route. Shopping or social trips may permit higher departure and arrival flexibility, and lower early and late arrival penalties. 2.1 Research framework Following the aforementioned concepts, this research proposes a behaviourally induced, system optimal model following the framework illustrated in Figure 2-1. The major body of this work flow is the iteration included in the blue box, which has five major components followed by a link flow update: 1) Marginal cost calculation / update 2) Time-dependent shortest path (TDSP) search 3) Schedule delay ( calculation 4) Optimal path finding 5) Traffic volume update

5 Figure 2-1: Research Framework

6 The detailed modelling procedure as described below starts from a pre-defined departure and route choice condition, which could be User Equilibrium (UE) or any other non-ue or non- SO condition: Step 1: Randomly select a certain percentage of vehicles from the dataset to be the experiment vehicles. Step 2: Read traffic demand for each link. Here the demand is calculated by scanning through the vehicle trajectory file and sum up all the vehicles on the same link at the same time. The reason for not using the link volume to approximate demand is that the existence of link capacity constraint in the peak hours might cause an underestimation of link demand. Step 3: Set i = 1 and compute the link marginal cost. Use the traffic demand calculated in Step 2 together with other available data to derive the marginal cost for each link; the equations used will be demonstrated in Section 3.2 in the full paper. Step 4: Set t = t 0, and find time-dependent minimal marginal cost path using the link marginal cost computed in Step 3, Section 3.3 in the full paper further describes the TDSP search algorithm in detail. Step 5: Compute the SD for that path (the SD calculation method will be presented in Section 3.4 in the full paper). Step 6: If there is another departure time choice for the user, update the departure time t and go back to Step 4; otherwise go to Step 7. Step 7: Among all different departure time and path choices, choose the one with the least general cost, which is computed through a linear combination of travel time and schedule delay. The detailed logic of finding the optimal path will be explained in Section 3.5 in the full paper. Step 8: Assign that optimal path to user i, and update the link volumes of interest, i.e., decrease the link s volume along the previous route by 1, and increase the link s volume along the new route by 1. Step 9: Determine if all experiment vehicles have been assigned a new path; if yes, go to Step 10. Otherwise, set i = i + 1 and go back to Step 3 to iterate, until all experiment vehicles have found a new optimal route Step 10: Re-write the simulation input file using the vehicles' new path and/or departure time information, feed back into the simulation system

7 Step 11: Re-run the simulation for one shot using the updated vehicle and path information as input, and analyze system benefit as well as other stats. 3 CASE STUDY The case study to illustrate the system performance and behavior changes of the proposed SO model is part of the report named Mobility Investment Priorities Project: Long-Term Central Texas IH35 Improvement Scenarios conducted by Texas A&M Transportation Institute (TTI), Texas A&M University (TAMU) System and College Station, TX, in preparation for Texas Transportation Commission and 83 rd Texas Legislature in Aug. 2013, in which report it was built off the core scenario (SC2) as supplemental analysis (SC2d). (Jeff Shelton et al. 2013). 3.1 Background The City of Austin is among the fastest-growing cities in the U.S.. Travel times from downtown Austin to Round Rock, where many commuters live, range from 45 to 60 minutes during the average afternoon rush hour. And yet, there is no agreement on what should be done to solve the travel time problem. The long-range transportation plan for the Austin Metropolitan Area, the 2035 CAMPO Metropolitan Transportation Plan (2035 CAMPO), shows no large-scale construction improvement strategies for IH 35 through Central Texas. Ongoing IH 35 initiatives by the Texas Department of Transportation (TxDOT) and the City of Austin focus on short- and mid-term improvement strategies that address existing and near-term congestion with potential high-return strategies. At the same time, decision makers have expressed a need for examination of long-term solutions for IH 35, considering, for example, concepts which had been discussed under previous studies but not fully explored. The following table summarizes the initial scenarios considered in the report.

8 Table 3-1 A list of initial scenarios tested In addition to the 7 core scenarios, the above proposed behaviorally-induced System Optimal Approach is listed as part of the Technology Strategy that was built off Scenario 2 (SC2) as supplemental study as Scenario 2d (SC2d). Several levels of driver market penetration of the

9 SO routing application were tested (5%, 10%, and 20%), along with several levels of allowable flexibility (15, 30, and 60 minutes) to vary departure time from that specified in the original origindestination trip table used for Scenario 0, the base case. 3.2 Travel time saving capability To test the travel time saving capability of the proposed methodology, the parameters below were used. The result shows the proposed approach is able to save 215,900 hours travel time per day for the system as a whole and is one of most efficient approaches to reduce traffic congestion. SO Routing used only during peak period. 20% driver market share, that is, drivers using the application. 60-minute maximum flexibility that drivers are allowed to deviate from their original departure time (not all users will use all 60 minutes). 3.3 Technical Assessment Heat diagrams were generated and used in this research to show the temporal-spatial speed profile of the IH 35 corridor in the case study network. The X-axis stands for the time and the Y- axis represents the physical road segment along the corridor; the color in the heat diagrams describes the congestion levels in the traffic network, with blue as free flow condition and red as severely congested. Figure 3-1 compares Scenario 2d assuming the SO routing technology to Scenario 0. As shown, there is a noticeable improvement in both the northbound and southbound directions under Scenario 2d. It demonstrates the potential of an existing application for providing measurable impact to reduce IH 35 general purpose lane congestion, even under the extremely congested circumstances projected for the year 2035.

10 Figure 3-1 Scenario 2d Heat Diagram Comparison to Scenario Technology Strategy Sensitivity Testing Table 3-2 lists various sensitivity tests that were conducted to assess the impact of different percentages of drivers using the technology application, as well as that of the flexibility allowed (not necessarily utilized) by drivers when they departed for their vehicle trip. Table 3-2 demonstrates that with the proposed approach, SO Routing users, experience a benefit as a result of using the technology themselves, and as more users participate in using the technology and

11 being more flexible with regard to the departure, more benefit in terms of travel time saving can be observed. Table 3-2 Technology Strategy Sensitivity Testing 4 CONCLUDING REMARKS This paper documents the research effort in developing a behaviourally induced, SO model to improve the system performance towards the SO condition. Both analytical derivation and numerical analysis have been conducted on CAMPO network in Austin, TX, as part of the mobility investment priorities analysis report prepared for Texas Transportation Commission and 83 rd Texas Legislature. The outcome of this study shows that our proposed method offers a promising modelling approach for improving the existing network traffic condition towards SO, and that the amount of economic value that can be saved by the proposed system can be significantly huge.

12 5 REFERENCES Jeff Shelton, Karen Lorenzini, Gabriel (Alex) Valdez, Williams, T. and Crum, S. (2013). Establishing Mobility Investment Priorities Under TxDOT Rider 42: Long-Term Central Texas IH 35 Improvement Scenarios. Prepared for Texas Transportation Commission And 83rd Texas Legislature. Austin, TX, Texas A&M Transportation Institute, The Texas A&M University System, College Station, Texas