Login

Project

#305 Data-driven mobility service design: a case study for Moon Township


Principal Investigator
Sean Qian
Status
Completed
Start Date
July 1, 2019
End Date
Aug. 30, 2020
Research Type
Applied
Grant Type
Research
Grant Program
Private Funding
Grant Cycle
2019 Smart Mobility Challenge
Visibility
Public

Abstract

This project investigates the characteristics in demand from existing last-mile microtransit operations and proposes novel methods for data-driven transit system design provided with spatially sparse demand (in suburban areas). The scope of this study is the area of Moon Township in Pittsburgh Metro Area. This research requires simulating operations of a feeder microtransit service in a medium-sized area based on known and predicted demand. The model developed in this research is aimed to be generic and hence aimed to be applicable for similar feeder services in other places. In the near future, the model and planning strategy can be further calibrated with more data and support operations managements decisions for multiple stakeholders such as public transit operators, community planners, microtransit operators, etc.

The proposed research project will: 1) Develop a generic demand model for sparse microtransit demand to estimate/predict the time-varying passenger demands in Robinson Township. The model estimates the presence/quantum of demand between depots and different destinations in the first-mile/last-mile operations of RideACTA (in particular their time-varying demand characteristics); The current microtransit service will be simulated to understand the demand response and service efficiency; 2) Create a feeder service transit model for Moon township area based on estimated demand; 3) Simulate operations in Moon Township with estimated/predicted demand and estimate the effectiveness of the proposed operation model through the simulation. 
    
Description
Tasks

Task 1: Create an estimation/prediction model for Robinson Township

Obtain the ridership data from RideACTA, which contains the following fields needs to be preprocessed:
•	origin and destination locations of rides
•	origin and destination timestamps of rides
•	passengers riding on the shuttle
•	passenger waiting times
Create a model which can estimate or predict the presence or quantum of demand for RideACTA’s microtransit service. The accuracy of the model is to be determined by splitting the data between training and test data and estimating the quality of predictions over the testing dataset using appropriate measures. 

Provided with the demand, the current microtransit service will be simulated to understand the demand response and service efficiency.

Deliverable:
•	Provide a report on modelling efforts to include a description of data used, establishing baselines, description of final method and results of numerical experiments to validate the model

The initial preprocessing and the model development is completed is expected to be completed in August, 2019 (2 months from start of project). 

Task 2:  Establishing a simulation model for ridership in Moon Township

This task requires creating a framework for a simulation with variables to track, such as total/average delay for passengers, total/average waiting time, total vehicle miles travelled, number of shuttles used etc., model the Moon Township area and the routes taken by the shuttles to fulfill demand, model time-varying demand for different times of operation and apply appropriate operational constraints.

In the simulation, the time-varying demand for Moon Township is calibrated through trends observed in Task 1. The operational constraints such as headway between vehicles, fixed or flex-routes, fixed or flex-stops, is found by assuming an initial demand and then finetuning from simulation results. The routing is done by using appropriate Vehicle Routing Algorithms. The results of the simulations should give the appropriate variables of interest such as those suggested in the beginning of this task.

Deliverable: 
•	Provide a detailed report on modeling efforts and findings.
The modelling of demand levels, identification of appropriate operational constraints based on demand levels, and setting up the simulation is expected to finish in February, 2020 (a total of 6 months after the end of previous task).

Task 3: Proposal of a transit system and the evaluation of the proposed transit system using simulation
Different demand scenarios are tested out in this phase using the simulation model created in Task 2. The operational constraints, which constitute the proposed transit system design, are identified and finetuned based on observations of the simulations. Also, this task includes identifying data-driven decision-making rules for change in operational constraints using the simulation model. 

In addition, for a low demand-level scenario, we study the possibility of using different travel modes to complement the last mile service. Those modes include TNC companies, such as Uber/Lyft/Ztrip, or shared bikes/scooters. We will provide a cost-benefit analysis for running the designed transit system under a low demand level, with and without complementary travel modes.  

Deliverables:
•	Provide a set of recommendations for transit system operation based on expected demand levels
•	Provide a report on modeling efforts and findings
The expected completion of this part is May, 2020 (3 months after the completion of previous task).

Task 4:  Draft Final Report  
Upon completion of Tasks 1-3, a draft final report will be generated that summarizes all findings and provided to RideACTA and Traffic21 for review.

Deliverable:
•	Provide a draft final report summarizing all finding from this research. 
The expected completion of this part is June, 2020 (1 month after the completion of previous task) and hence concluding the project.


Based on time constraints, the following optional tasks may be considered.

Task 5:  Create an application for ride management for RideACTA  
A web-based application which helps coordinate the operations of RideACTA (or similar microtransit operators) by providing interfaces to riders, drivers and operation managers is proposed. The proposed web-application is supposed to use real-time predictive analytics to help optimize operation costs for RideACTA and provide a higher level of service to its riders.
    
Timeline
07/01/2019 - 06/30/2020    
Deployment Plan
We wil work with RideACTA closely to make decisions on optimal shuttle services, and provide RideACTA all materials to run trials in the field.    
Expected Accomplishments and Metrics
•	Provide a set of recommendations for transit system operation based on expected demand levels
•	Provide a report on modeling efforts and findings for RideACTA

Metrics: shuttle service efficiency and costs, riders' waiting time and the number of unmet demand. 
    

Individuals Involved

Email Name Affiliation Role Position
mbattifa@andrew.cmu.edu Battifarano, Matthew CMU Other Student - PhD
bethannh@andrew.cmu.edu Hockenberry, Beth Carnegie Mellon University Other Other
seanqian@cmu.edu Qian, Sean CMU PI Faculty - Untenured, Tenure Track
Pengjiz@andrew.cmu.edu Zhang, Pengji CMU Other Student - PhD

Budget

Amount of UTC Funds Awarded
$85887.45
Total Project Budget (from all funding sources)
$135394.00

Documents

Type Name Uploaded
Data Management Plan dmp_ACTA.docx May 16, 2019, 11:39 a.m.
Progress Report 305_Progress_Report_2019-09-30 Sept. 23, 2019, 8:33 p.m.
Progress Report 305_Progress_Report_2020-03-31 March 23, 2020, 9:39 p.m.
Final Report Final_Report_-_305.PDF Sept. 1, 2020, 5:49 a.m.

Match Sources

No match sources!

Partners

Name Type
RideACTA 1