2021

Pablo Martín Calvo, Universiteit van Amsterdam
Bas Schotten, Municipality of Amsterdam
Elenna Dugundji, Vrije Universiteit, Amsterdam

On-street parking policies have a huge impact on the social welfare of citizens. Accurate parking occupancy data across time and space is required to properly set such policies. Different imputation and forecasting models are required to obtain this data in cities that use probe vehicle measurements, such as Amsterdam. In this paper, the usage of traffic data as an explanatory variable is assessed as a potential improvement to existing parking occupancy prediction models. Traffic counts were obtained from 164 traffic cameras throughout the city. Existing models for predicting parking occupancy were reproduced in experiments with and without traffic data, and their performance was compared. Results indicate that: (i) traffic data is indeed a useful predictor and improves performance of existing models; (ii) performance does not improve linearly with an increase in the number of counting points; and (iii) placement of the cameras does not have a significant impact on performance. Keywords: parking occupancy, parking prediction, sparse spatial-temporal data, traffic counts, gradient boosting machine

Ye Zhang, Beijing University of Technology
Yanyan Chen, Beijing University of Technology
Haodong Sun, Beijing University of Technology

With the improvement of living standard and the rapid development of transportation industry, tourism travel has become a typical travel behavior. However, the increase in tourist population has led to a surge and detention in urban passenger flow, resulting in insufficient urban transportation capacity and traffic congestion. To solve this problem, it is important to accurately obtain the characteristics of the tourist flow and understanding their changing rules. And it has been gradually recognized that mobile phone can be used as a practical and promising way to identify individual travel trajectories and extract travel characteristics of the group. Therefore, this paper proposes a method to obtain the characteristics of tourists based on mobile signaling data. Firstly, we calculated the accessibility and connectivity of tourist attractions in Beijing and chose the attraction which has the worst accessibility as a case to analyze in the next work. Secondly, we extract the user's Origin-Destination (OD) trip through the pre-processed mobile signaling data. Thirdly, we identify the tourists who arrive at the tourist attraction and describe their spatial-temporal characteristics. Finally, the results can provide decision-making support for the tourism, traffic management departments and tourists, and improve the service level of tourism traffic.

Jason Hawkins (jason.hawkins@mail.utoronto.ca), University of Toronto
Khandker Nurul Habib, University of Toronto

Data are important components of any research; however, it is often the case that the required data are not readily available. Researchers must often fuse multiple datasets to obtain the data required to complete their work. In some cases, researchers must also synthesize the necessary data. In this paper, we address both these problems and propose a data fusion framework for constructing a statistically valid synthetic population for use in urban simulation models. We proceed from the fusion of datasets, to the synthesis of individual and household characteristics, and the expansion of these synthetic data into a full population. We develop the framework for the case of constructing a dataset representing patterns of household-level expenditure and individuals’ time-uses. The Greater Toronto Area (GTA) in Canada is used a testbed for the method. The results of the data fusion and synthesis are validated against statistics from a large-sample travel survey conducted in the GTA, showing a good fit with the validation dataset. Finally, we outline how the framework could be applied in other contexts where a single dataset is unavailable.

yuan tian, University of Nevada, Reno
Yuxin Tian, University of Nevada, Reno
Hao Xu, University of Nevada, Reno
Trevor Whitley, University of Nevada, Reno
Shradha Toshniwal, University of Nevada, Reno

Traffic volume, including lane-based traffic volume and pedestrian (bicycle) crossing volume, is critical for traffic planning, design, and operation. Traditional volume evaluation methods mainly depend on loop detectors which only provide simple macro traffic information and cameras that require adequate light conditions. With the latest innovative uses of LiDAR sensors, road user’s trajectory obtained from LiDAR is becoming more attractive. However, since trajectory data is still defective because of the occlusion issue and limitation of the detection range of the sensor, there is no existing method to extract volume using it. To get more accurate traffic volume, an automatic best detection zone searching method is provided. In this method, the input is the trajectories from roadside LiDAR data. The first step is to determine the boundaries of the searching area. Then, a traversal search is applied to find the detection zone for each lane and crosswalk. To reduce the searching time, the true volume that will be used for comparison to decide when to stop searching is calculated. When the best detection zone is available, large-scale data and real-time traffic volume can be processed. A whole day of data collected in the real world is selected to verify the method, and the result shows that the accuracy of this traffic volume extraction method reaches 95% or higher. Traffic volume extraction with processed roadside LiDAR data will significantly change how traffic agencies assess road network performance and add great traffic values to the existing probe-vehicle data and crowd-resourced data.

Hima Shaji, Indian Institute of Technology, Madras
Lelitha Vanajakshi (lelitha@iitm.ac.in), Indian Institute of Technology, Madras
Arun Tangirala, Indian Institute of Technology, Madras

A possible direction towards the development of sustainable transportation solutions is to improve the quality of public transportation and attracting more travellers. This requires accurate prediction of bus travel times to improve the quality of public transport. Drawing meaningful inferences from the data and using these to aid in prediction tasks is always an area of interest. Earlier studies predicted bus travel times by identifying significant regressors, which were identified offline based on chronological factors. However, travel times vary both spatially and temporally. Moreover, these patterns may not be static and may vary depending on factors such as day, time, and location. A related question is whether the prediction accuracy can be improved with the choice of input variables being used in a given prediction algorithm. The present study analyses this question systematically by presenting the input data in different ways to the prediction algorithm, such as presenting the data without grouping, dividing the dataset manually based on chronological factors, and using a data-driven technique to divide the dataset into groups with similar features. The focus of this study is on whether using such grouping information can improve the final predictions. It was observed that the prediction accuracy increased when the dataset was grouped and separate models were trained on them, the highest accurate case being the one where the dataset was divided using the data-driven automated technique. This study demonstrates that understanding patterns and/or groups within the dataset helps in improving the accuracy of predictions.

Sneha Roy, AECOM
Drew Cooper, San Francisco County Transportation Authority (SFCTA)
Richard Mucci, University of Kentucky
Bhargava Sana, San Francisco County Transportation Authority (SFCTA)
Mei Chen, Kentucky Transportation Center
Joe Castiglione, San Francisco County Transportation Authority (SFCTA)
Dr. Gregory Erhardt, University of Kentucky
Sneha Roy, AECOM

Traffic congestion has worsened noticeably in San Francisco (SF) and other major cities over the past few years. Part of this change could reasonably be explained by strong economic growth or other standard factors such as road and transit network changes. It also corresponds to the emergence of Transportation Network Companies (TNCs), such as Uber and Lyft, raising the question whether the two may be related. Our research incrementally decomposes the contributors to increased congestion in SF between 2010 and 2016, namely, road and transit network changes, population growth, employment growth, TNC-volumes, and the effect of TNC pick-ups and drop-offs. We do so through a series of controlled travel demand model runs, supplemented with observed TNC-data collected from the Application Programming Interfaces of Uber and Lyft. Our results show road and transit network changes over this period have only a small effect on congestion. Socioeconomic factors contribute about a quarter of the congestion increase, and that TNCs are the biggest contributor to growing congestion over this period, contributing about half of the increase in vehicle hours of delay, worsening travel-time reliability. This study is more data rich and spatially detailed than past studies, providing a better understanding of where and when TNCs add to congestion. It gives transportation planners a better understanding of the causes of growing congestion, allowing them to more effectively craft strategies to mitigate or adapt to it.

Yan Liu, University of Cincinnati
Adekunle Adebisi, University of Cincinnati
Tao Li, University of California, Los Angeles
Jiaqi Ma, University of California, Los Angeles

In the past decade, private sector data, as an emerging data source, has been gradually adopted by many public agencies for performance measurement and travel time provision, mostly on freeways. Agencies have also used data from infrastructure sensors and floating cars to validate the quality of the private sector data on both freeways and arterials under different conditions. While freeway data quality has been proved by many studies, non-freeway travel time data have been considered unreliable, particularly when the segment is congested and high-density signals exist. In recent years, data vendors started to provide high-resolution travel time data (e.g., INRIX XD data), which are provided on much shorter segments than conventional Traffic Message Channels (TMC). This study investigates the data quality of this new data source for providing travel times to non-freeway users. The study evaluates private sector data by using “floating car” and Wi-Fi travel time data as the ground truth. The results show that the new private-sector data on non-freeways are generally acceptable for segments with low and moderate congestion. With higher congestion, the data quality is site-dependent. A neural network (NN) model was proposed to correct travel time with private sector data. It is recommended that this new data should be validated for each site before being used for travel time provision or other purposes, and the data correction can be applied for sites with similar characteristics that impact private sector data accuracy.

Feilong Wang, University of Washington
Jingxing Wang, University of Washington
Yiran Zhang, University of Washington
Cynthia Chen, University of Washington
Xuegang (Jeff) Ban, University of Washington

The Alaskan Way Viaduct in Seattle, an elevated freeway, was replaced with the new SR 99 Tunnel in 2019. The project involved a sequence of events, including the closure of the viaduct for demolition, reopening (without tolling) and followed by tolling of the new tunnel that was built to replace the demolished viaduct. This study examines travelers’ adaptive behavior in response to these events. We used both big data (i.e., app-based data) and flow data (traffic volumes and travel times) to conduct before/after analyses and comparisons, and presented the data processing and analysis methods. Impacts of these events are analyzed in terms of changes in travelers’ mobility patterns and system performance at three scales: facility users, the immediate neighborhoods near the project site, and the entire Puget Sound region. Results show that the impacts are more significant on facility users who use the viaduct or tunnel regularly than those who live in neighborhoods nearby the facility; there is almost no impact at the regional scale. Meanwhile, the viaduct closure seems to have less significant impacts than the tunnel tolling, suggesting that the information dissemination and guidance provided by transportation agencies regarding the viaduct closure may have helped reduce the potentially larger impacts. The analysis method and results presented here also indicate that big data fused with traditional flow data can be useful for analyzing the impacts of major infrastructure changes, if properly conducted.

Qiuhan Chen, Georgia Institute of Technology (Georgia Tech)
Ariel Steele, Georgia Institute of Technology (Georgia Tech)
Yichang Tsai, Georgia Institute of Technology (Georgia Tech)

Complete streets ensure that all motorists, cyclists, transit riders, pedestrians, wheelchair users, and other travelers have equal access to transportation facilities. They improve safety, lower carbon emissions, encourage exercise, and enhance quality of life. To implement complete streets on a network level, transportation agencies are in the process of finding cost-effective methods for inventory and assessment. Efficient data collection is needed to support the development of a robust asset management plan. Thus, there is a need for a technology review. The objectives of this paper are to identify and categorize complete street features and attributes that are essential for complete street asset management, to comprehensively review the technologies that have been explored for complete street data collection, to critically assess and identify the challenges and needs of technologies for complete street data collection, their implementation status, and their strengths and limitations for specific complete street feature/attribute data collection, to survey the cost and maturity of implementation for the promising technologies, and to provide recommendations for pilot studies to improve data collection, data processing, and data analysis based on the current gaps. It was found that LiDAR, 3D laser, smart phones, and satellite imaging are some of the most promising technologies. Of the three categories of attributes (presence, quality, and utilization), it was found that measuring presence is easier than collecting data about quality and utilization, and thus those should be the key areas for development moving forward.