Authors: Shaghayegh Shabanian (Cambridge Systematics), Kenneth Curry and Matthew Austin (Parsons)

Abstract Title
Isolating the Effects of HOV Weaving: A Trajectory-Based Microsimulation Approach

Topic
Microsimulation and Big Data

Abstract Background
High-occupancy vehicle (HOV) lane weaving is often recognized as a significant contributor to freeway congestion, yet its isolated impact is rarely quantified due to confounding factors such as spillback from downstream bottlenecks, insufficient capacity at exit ramps, and other operational issues. To support an action plan by Caltrans to address degraded HOV lanes, this study systematically evaluates whether weaving maneuvers to and from continuous-access HOV lanes are primary causes of performance degradation on freeway segments. Advanced microsimulation techniques were utilized to quantify these impacts comprehensively. The study area was a 6-mile segment of I-5 in Irvine, California.

Description of Abstract
The study began with a pre-screening process to identify freeway segments most likely impacted by weaving. Using PeMS detector data, which provides lane-specific volume and speed, and preliminary Highway Capacity Manual (HCM)-based analysis, segments were evaluated for factors such as congestion patterns, ramp volumes, and the presence of direct HOV ramps. This ensured selection of a segment where weaving could be a primary issue.

A comparative evaluation of microsimulation tools was conducted, focusing on the ability to control lane-specific calibration, model continuous-access HOV lanes, and flexibility in lane change behavior.

The selected segment was modeled and meticulously calibrated in TransModeler. Calibration focused on replicating origin-destination patterns, lane-specific volumes, and speeds. Vehicle trajectories generated by the simulation were analyzed in detail to ensure accurate lane change behavior, a critical element in evaluating weaving impacts.

To isolate the weaving effect, modified scenarios were created by adding ramps on the left-hand side, enabling direct exits from the HOV lane and eliminating weaving maneuvers. Performance measures, including travel time and the number of lane changes were compared between the base and modified scenarios. This comparison quantified the direct impact of weaving on congestion and overall traffic performance.

The findings affirm the effectiveness of trajectory-based calibration in microsimulation modeling and the utility of the pre-screening process in identifying segments where weaving significantly impacts performance.

Statement on Why Abstract is Noteworthy
This work is noteworthy for its emphasis on detailed calibration and the measurement of weaving impacts through alternative analysis. Furthermore, it could guide the selection of cost-efficient treatments that deliver the greatest benefits in mitigating weaving-related congestion.

Project is Complete

Abstract Background

Roadway incidents are unplanned or planned events that cause disruption to regular traffic operation. Roadway incidents’ root cause can be linked to work zones, crashes, weather events, or planned special events. These types of incidents can significantly impact traffic performance causing increased delays, emissions, user cost, and reduced economic business revenues. Roadway incidents may vary spatially and temporally (e.g., short-term or long-term) eliciting varying traveler and route choice behaviors. Proper modeling and evaluation of these disruptions are essential for identifying bottlenecks, optimizing traffic diversion strategies, and improving network resilience. Interstate 10 (IH 10) traverses the state of Texas from east to west and serves a major corridor that supports economic and social development in the region. This study evaluates the impacts of incidents to the IH 10 corridor in the Greater Houston metropolitan area. Results from the modeling and assessment of roadway incidents will support short and long-term planning of the corridor to improve its resilience to recurrent and non-recurrent roadway incidents.

Description of Abstract

This presentation introduces an analytical framework designed to model and assess the impacts of roadway incidents. The study develops multiple performance measures, including:

• Vehicle Hours Traveled (VHT), Vehicle Miles Traveled (VMT), Total vehicle delay
• Passenger vehicle and truck user costs
• Travel time for typical and alternative routes to/from activity centers
• Availability of alternative routes to/from activity centers
• Accessibility to jobs and population from activity centers
• Truck travel time and alternative route availability between key interest areas

The study utilizes mesoscopic dynamic traffic assignment model, DynusT. This model captures user behaviors based on historical, en-route, and pre-trip information, allowing for detailed trip tracking by vehicle type and user behavior. DynusT supports the modeling of both planned and unplanned events through different assignment methodologies including one-shot run or full Dynamic User Equilibrium (DUE). 

The study evaluates eight hypothetical incident scenarios along IH 10 in the Houston area, as depicted in the figure (attachment). Scenarios S1 to S5 involve freeway closures at specific spot locations, whereas scenarios S6 to S8 involve closures within defined roadway segments. Scenarios S1 to S4 are characterized as unplanned, short-term events. Conversely, scenarios S5 to S8 are modeled as planned, long-term events. Results are compared to baseline scenario representing regular traffic operations. 

Statement on Why Abstract is Noteworthy

This study offers a comprehensive, transferable, and practical framework for assessing roadway incidents and provides practitioners and stakeholders with data-driven information for informed decisions to enhance system resilience. Through an accurate modeling of roadway incidents, the study identifies bottlenecks, evaluates traffic diversion effectiveness, and formulates strategies to improve mobility under disruptive conditions. 

Study Status

 This study is expected to reach significant milestones by Fall 2025:

●      Scenario modeling completed by February 2025

●      Scenario analysis completed by March 2025

Other study tasks will extend beyond Fall 2025. The findings from this study will provide critical inputs for ongoing planning efforts.

Application of Distributed Value of Time (VOT) in Agent-Based Dynamic Traffic Assignment (AB-DTA) for Toll Road Forecasting
J. Hicks, P. Vovsha, M.Mahut, I.Juran (Bentley), H. Zhu, A. Dutta (MAG) 
Word count: 473
The paper reports on the results of a detailed test of AB-DTA and evaluation of the impact of different assumptions on the distribution of VOT.  In the AB-DTA route choices are individualized allowing each vehicle to have its own VOT.  Tests were conducted for a large-scale real-world network of the Phoenix metro area with a hypothetical highway pricing project using fixed demand over a wide range of toll values. 
Based on prior research on heterogeneity of users (Jiang & Mahmassani, Tian & Chiu), preliminary expectations were established regarding the relative number of toll users as a function of the toll value with the mean VOT by class and continuously distributed VOT.  The continuously distributed VOT followed a theoretically established lognormal distribution with the same mean value for each class as the fixed VOT.  The individualized VOTs were assigned in the Activity-Based demand model (ABM) as a function of trip purpose, person income, and car occupancy and corresponded to the VOT used in the mode and other travel choices.    
Two sets of tests were implemented. In the first set of (simpler) tests, the demand was generated by the ABM and kept fixed assuming that only routes change with toll values. In the second set of (more complex) tests the integrated ABM-DTA model was applied allowing for a more complete response including demand as well as route choice.  The implemented tests provided important insights with recommendations for toll road traffic and revenue forecasting:
•    There is a substantial difference between the mean VOT and continuously distributed VOT results, indicating that VOT setting (distributed VOT vs. constant class-based VOT) is an important factor that must be considered carefully.  In general, continuously distributed VOT is more realistic, and use of this VOT setting and corresponding AB-DTA should be utilized going forward as the main tool for analysis of toll scenarios.      
•    For a given value of mean VOT, ignoring the VOT distribution results in a systematic overestimation of toll traffic and revenue for lower toll values and systematic underestimation for higher toll values.  For lower tolls, the mean VOT exhibits a higher volume of toll users since only the continuously distributed VOT scenario can contain vehicles with very low VOT that would respond to these levels. For higher tolls, the continuously distributed VOT exhibits a higher volume of toll users since only the continuously distributed VOT scenario can contain vehicles with very high VOT that would not be affected at these levels.  
•    The demonstrated magnitude of these systematic biases warrants a word of caution for utilizing simplified VOT settings in practice. 
•    Given the fact that advanced ABMs today are characterized by a detailed segmentation of travel (and can naturally incorporate continuously distributed VOT within each segment), AB-DTA represents the best network model for an integration with ABM that preserves a full consistency of individual VOT between these two models.   
 

This presentation describes a recently completed full-day application of DTA for the metropolitan region of Phoenix, AZ as part of the advanced demand forecasting project integrating activity-based demand with dynamic traffic assignment models (ABM-DTA). The full-day traffic simulation of 20M vehicle trips in a large network represented a challenge where a broad range of factors and desired properties comes into play such as ensuring high levels of model convergence and practical run times. Traditionally, run times are a barrier to growing these applications to metropolitan and regional scales.

In contrast to typical multiresolution approaches that combine entirely different modeling paradigms (macro/meso/micro) in a discontinuous way across the network, the approach described here is based on a single underlying traffic model, which is a vehicle-based mesoscopic simulator, that can be solved mathematically at different levels of resolution. The key advantage is that for discrete simulation models, lower resolution allows for a major reduction in run time. At the same time, using a single traffic model ensures that the underlying model properties regarrding the representation and realism of traffic flow, and thus model results, are similar across different levels of resolution. This approach is similar to the idea of simulating packets, which was popular in some of the early work on flow-based mesoscopic models. This requires bucket-rounding the demand matrices, or, in the ABM-DTA context, extracting an unbiased sampling of households, which is a common practise for ABM.

We investigated the results of this approach used within an integrated ABM-DTA model, applied to a long-term future scenario for the regional model of Phoenix, AZ (USA), for assessing the impacts of managed lane policies on the freeway network. Multiple tests with different sampling and convergence protocols were implemented during the project where the results with sampling (low resolution) were compared to full sample (high resolution) model runs.  We’ll present some key results along with the savings in run time obtained with the multiresolution model.

The proposed approach to multiresolution mesoscopic modeling is shown to greatly extend the practical size of DTA models, to metropolitan-regional scales, thus reducing the need for conventional multiresolution approaches that combine fundamentally different modeling paradigms. Moreover, for ABM-DTA, this approach opens a way to implement a meaningful level-of-service feedback in the integrated model at the individual level.

A discussion with the session panel and audience about challenges and opportunities associated with utilizing mesoscopic simulations (DTA) in planning applications.