Abstract
We identified two key operational issues faced by the Heritage Community Transportation (HCT): high cost, and long rider wait times. We propose adding flexibility in HCT’s fleet portfolio and routes to simultaneously address such issues. The project will collect data on demand patterns, develop new theory, and provide practical policy recommendations. It also provides faculty and students the opportunity to conduct research with a transportation provider whose mission is to serve our region’s most vulnerable populations.
Description
(I) Problem Identification:
Heritage Community Transportation (HCT) proposed a transportation problem via the Smart Mobility Challenge in October 2020. The problem was outlined as follows:
“COVID-19 has reduced the use of public transportation. Despite the relatively low risk of COVID-19 transmission on public transit, implementation of CDC guidelines and publicly available information how passengers can reduce risk of transmission, public transit ridership is down. Rising operating costs with no increase in annual funding, road and bridge construction projects and reduced passenger capacity due to social distancing have made serving the public increasingly difficult. HCT's cost per passenger increased from $12.94 to $24.04 in under six months. We propose a project to analyze HCT service data and public data to ensure that our vans are utilizing the best available routes and operating in the most cost-effective, rider-focused manner. Although currently operating as a fixed route line, HTC is also interested analyzing if alternative public transportation methods (deviated fixed route, on-demand pickups, autonomous vehicles, etc.) could better serve our riders.”
We think this problem can be scoped into an applied research project in operations/transportation, with very strong positive impact. This project would provide us the opportunity to conduct both transportation and socioeconomic research, with the goal of serving the most vulnerable populations.
In particular, we identified two key issues (via preliminary data and conversations with HCT; see “Additional Background Information” below) in the operational problem faced by HCT: high cost (for HCT) and long wait times (for travelers). These two issues, even though proposed via a COVID-related problem statement, are actually their long-term problems.
(II) Proposed Solution Framework:
We postulate that there is an opportunity to reduce HCT operational cost and reduce traveler wait times at the same time. Currently, they have three routes, and one van serving each route continuously from early morning to late evening. Often times the vans are not full, and at the same time, travelers would have to wait for the van for one or two hours for it to loop back if they miss it. The opportunity lies in a more flexible transportation plan, which includes two dimensions of flexibility:
1. Routing flexibility
a. Current: fixed routes
b. More flexible option: deviated fixed routes
c. Most flexible option: on-demand pickups
2. Fleet flexibility
a. Current: one van for each route
b. More flexible option: portfolio of small and large vehicles
c. Most flexible option: portfolio of vehicles, and travel reimbursement if use other transportation providers such as TNC
We think by adding some levels of flexibility in these two dimensions into the transportation system can help HCT reduce cost and improve rider experience at the same time.
(III) Challenges and Research Opportunities
It is well-known in the operations management and manufacturing literature that flexible resources can add value and reduce cost. However, it is not well studied in transportation theory, and it is unclear whether flexible routes and flexible vehicle types can provide sufficient cost saving and rider time saving in this kind of context. Therefore, even the deterministic version of this problem is a new and interesting theoretical question to study, and the analysis can provide practical insights on the upper bound of the benefit of flexibility.
We note that even though the analysis has not been studied extensively in theory, there are examples of successful implementations in practice: the Ride ACTA project led by Professor Sean Qian and his team, as well as our independent conversations with commercial transportation providers such as HopSkipDrive in other research projects.
One main challenge for taking this deterministic, theoretical analysis into a practical, robust solution is that there is no understanding of the actual travel demand information at the time. HCT collect travel data via surveys and detailed ridership timestamps at each pickup location, but this is endogenous on the fact that riders know van schedules and adapt to the van schedules. This information does not necessarily reflect the true demand pattern. In addition, uncovering such demand pattern is also difficult, for two reasons: (1) it is not easy to survey every person for their trip demand, and (2) demand pattern changes over time. Therefore, a significant portion of the analysis should be done to understand the potential uncertainty in demand pattern, and how to provide a robust transportation solution to hedge against uncertainty.
At the moment, HCT has agreed to provide accurate demand information surveys in the near future, should this project move forward.
(IV) Additional Background Information:
We met with the Brandon Mahler (Director of Transportation) and Melanie Young (Manager of Transportation) of Heritage Community Initiatives via Zoom on Nov 17, 2020 to discuss the problem. Correspondences also followed to give the researchers more insight about their operations, via their 2018 and 2019 survey summary results. In particular, these documents were shared and found to be very useful for understanding the problem:
• 2018 Rider Survey Summary Results
• 2019 Rider Survey Summary Results
• 2020-11-12 Daily Shuttle report (one row of data as a sample). This is an example of the daily data we collect from our vans. Stop location, passenger on/offs/ geolocation (latitude and longitude) etc.
Overall, HCT states that the following data are readily available:
1. Real time Port Authority of Allegheny County Service Data.
2. HCT Real time Vehicle operations data and performance measurements (automatic vehicle location via GPS, cost per passenger, daily ridership, passenger on/offs, method of fares, odometer, actual vehicle hours, actual vehicle miles, actual vehicle revenue hours, actual vehicle revenue miles, routes, layovers, on/off road routing, passenger pass-ups, etc.)
3. Sociodemographic and socioeconomic data
Brandon also mentioned via email that should it be useful, HCT can conduct surveys in the future to quantify travel demand (time and location).
The HCT website also provides real-time van location information.
Timeline
Month 1 to Month 12:
M1-M2 (July, August 2021): Problem definition, literature survey, and survey design
M3-M6 (Sep-Dec 2021): Data collection and data mining with HCT
M3-M6 (Sep-Dec 2021): Deterministic model analysis
M7-M10 (Jan-April 2022): Uncertainty analysis, and robust modeling
M11-M12 (May, June 2022): Propose pilot plan for one route
Strategic Description / RD&T
Deployment Plan
We plan to engage with HCT closely in two different phases. The first phase is at the beginning of the project, where surveys need to be designed and distributed. This survey aims to uncover qualitative and quantitative information related to individual level travel demand over time, supplementing the current “censored demand” information collected via HCT’s current surveys and trip logs. This requires extensive communication between the research team and HCT (and potentially Port Authority), therefore will be done over four months from Sep to Dec 2021 (this gives us a good amount of time to concurrently develop theoretical analysis for the optimization model with flexible resources and flexible times during this period).
A second engagement comes at the last four months of the project (March-June 2021), when models become relatively mature, but require additional information and practical guidance to position the it as an implementable and systematic solution. We plan to incorporate such feedbacks and propose a solution as pilot for HCT to test run in one of their three routes. If successful, it will be natural to scale to their other operations in future projects.
Expected Outcomes/Impacts
1. Design and implement data collection plans for the various components of demand, shedding light on the time, intensity, and non-stationarity of demand patterns. This is useful for HCT even beyond our proposed project. Data is to be collected and stored by HCT, and shared to CMU in a secure and private way.
2. Understanding of the key technical, financial, strategic, and policy issues underlying the problem.
3. Theoretical analysis of the benefit of flexible resources (vehicle portfolio) and flexible routes in the transportation context.
4. Understanding of uncertainty, and how a robust solution framework should behave,
5. Pilot plan to test the feasibility of flexible last-mile transportation for riders in the Heritage Community Transportation service area.
Expected Outputs
TRID
Individuals Involved
| Email |
Name |
Affiliation |
Role |
Position |
| pyzhang@cmu.edu |
Zhang, Peter |
Carnegie Mellon University |
PI |
Faculty - Untenured, Tenure Track |
Budget
Amount of UTC Funds Awarded
$90000.00
Total Project Budget (from all funding sources)
$180000.00
Documents
| Type |
Name |
Uploaded |
| Data Management Plan |
DMP_BWVdco3.pdf |
Dec. 16, 2020, 1:53 p.m. |
| Project Brief |
Slides_for_Mobility21_368_Demand_learning_and_supply_optimization_for_last-mile_transportation.pptx |
March 19, 2021, 11:05 a.m. |
| Publication |
Mobility_Fairness_Dataset.pdf |
Sept. 30, 2021, 5:26 p.m. |
| Presentation |
Optimality Criteria of Constant and Affine Policies in Adjustable Robust Optimization |
Sept. 30, 2021, 5:26 p.m. |
| Progress Report |
368_Progress_Report_2021-09-30 |
Sept. 30, 2021, 5:28 p.m. |
| Publication |
Fleet Sizing and Allocation for On-demand Last-Mile Transportation Systems |
Oct. 24, 2021, 8:46 p.m. |
| Publication |
Adjustability in Robust Linear Optimization |
March 27, 2022, 11:49 a.m. |
| Publication |
Model Mis-specification and Algorithmic Bias |
March 27, 2022, 11:49 a.m. |
| Presentation |
Adjustability in Robust Linear Optimization |
March 27, 2022, 11:49 a.m. |
| Presentation |
Demand learning and supply optimization for last mile transportation in low-income neighborhood |
March 27, 2022, 11:53 a.m. |
| Presentation |
Adjustability in Robust Linear Optimization |
March 27, 2022, 11:53 a.m. |
| Presentation |
Demand learning and supply optimization for last mile transportation in low-income neighborhood |
March 27, 2022, 11:53 a.m. |
| Presentation |
Last mile transportation efficiency as social services |
March 27, 2022, 11:53 a.m. |
| Progress Report |
368_Progress_Report_2022-03-30 |
March 27, 2022, 11:53 a.m. |
| Publication |
Structural Characteristics of Equitable and Efficient Distributions |
May 2, 2022, 9:54 a.m. |
| Publication |
Joint Optimization of School Bus Routes and Last-mile Services |
May 2, 2022, 9:55 a.m. |
| Publication |
Mobility21 Final Research Report: Joint Optimization of School Bus Routes and Last-mile Services |
May 2, 2022, 9:55 a.m. |
| Final Report |
368_-_Final_Report.pdf |
Aug. 8, 2022, 4:52 a.m. |
Match Sources
No match sources!
Partners
| Name |
Type |
| Heritage Community Transportation |
Deployment & Equity Partner Deployment & Equity Partner |
