Summary:This study is sponsored by Amazon, Bellevue Transportation department, Challenge Seattle, King County Metro, Seattle Department of Transportation, Sound Transit, and Uber, with support from the Mobility Innovation Center at UW CoMotion.
Population and extended economic growth in many Seattle neighborhoods are driving increased demand for private car travel along with transportation services such as ridehailing and on-demand delivery. Together, these trends are adding to existing demand for loading and unloading operations throughout the city, and exacerbating traffic congestion. Anecdotal evidence indicates that passenger/delivery vehicle stops at or next to transit stops can interfere with bus operations, causing longer or more volatile delays. The increased travel times and reduced reliability further erode the attractiveness of transit to travelers. Thus, it is important to understand how transit, ridehailing, and goods delivery vehicles interact in terms of both operations and travel demand.
This project focuses on the analysis of open-source transit data to screen for locations with slow and/or unreliable bus travel times, and couple that data with interference observation, environmental, and traffic-related data to potentially predict the likely causes. We have developed tools to identify transit corridors with high levels of interference from other road users, including passenger cars, ridehailing vehicles and goods delivery vehicles. These tools are applied to transit corridors in Seattle and Bellevue, and methods have been developed to identify likely sources of interference from available data.
We drew on multiple data sources for identifying high-interference corridors in the region, including:
- a virtual workshop with participants from beneficiary agencies and stakeholders to solicit input;
- an online crowdsourcing survey to engage the community and gather feedback from all road users;
- route-level ridership data from King County Metro; and
- aggregated pick-up/drop-off data on ridehailing activities from SharedStreets.
Data was consolidated and 10 corridors were selected based on their likelihood of containing interference between buses and other road users, transit ridership levels, and stakeholder and community feedback.
In addition, we have developed a tool for identifying corridors with slow and/or unreliable bus travel times from open-source real-time transit data. We implemented a pipeline for ingesting and analyzing King County Metro’s real-time Generalized Transit Feed Specification data (GTFS-RT) at 10-second intervals. Using this pipeline, active bus coordinate and schedule adherence data has been scraped and stored to an Amazon Web Services (AWS) server since September 2020. We developed efficient methods to aggregate tracked bus locations and assign them to roadway segments, and quantified delays in terms of schedule deviation and ratio of median to free-flow speeds, among other metrics. We have developed a web based visualization tool to display this data, and it is being updated daily with aggregated performance metrics from our database.
To collect ground truth validation data along selected corridors, we implemented an online data collection tool for field observations, and recruited research assistants to observe bus operations along the study corridors and record information on bus traversals and instances of interference. This dataset is analyzed alongside the GTFS-RT data, environmental, and traffic related data to identify instances of delay and predict the likely causes.
Field data was collected for three weeks along eight of the selected corridors in March 2021, but was later paused due to depressed levels of transportation activity during the COVID-19 pandemic and the current unstable condition of travel choices and city traffic (and thus interferences). Preliminary analysis on the collected data revealed that there is not a substantial effect shown in the GTFS-RT data when a bus is interfered with; however, there were not a lot of interference observations in the collected field data. So, it remains to be seen whether the lack of an identifiable effect is due to the lack of ground truth data, lack of precision in the automatic vehicle location system, or the relatively low impact of an interference when compared to the effects of general traffic congestion, signals, and other roadway conditions. A linear regression model was also generated to determine the extent to which roadway characteristics can predict segment performance, which produced mildly predictive results.
As businesses and transit services continue to reopen, there will likely be an increase in the amount of transit interference experienced between buses and other roadway users, which will potentially allow for the gathering of more ground truth validation data. Field observations will resume in late Summer/early Fall 2021 and will continue until enough data is collected to either (1) model connections between observed interference and bus delays in the GTFS-RT data; or (2) determine whether significant delays cannot be linked to observed instances of interference in the study corridors. The GTFS-RT data scraping will continue daily, and summarized in the developed interactive visualization tool.
The major anticipated benefits of the project can be summarized as follows:
- This work will help identify network-wide road and route segments with slow and/or unreliable bus travel times. We may also be able to identify main causes of delay in the study corridors.
- Moreover, we expect that this work will generate reusable analytical tools that can be applied by local agencies on an ongoing basis, and by other researchers and transportation agencies in their own jurisdictions.
- The outcomes of this work will enable identifying corridors with slow and/or unreliable bus travel times as candidates for specific countermeasures to increase transit performance, such as increased enforcement, modified curb use rules, or preferential bus or street use treatments. Targeting such countermeasures towards priority locations will result in faster and more reliable bus operations, and a more efficient transportation network at a lower cost to transit agencies.