Embedded Files
Collaborating with University of Michigan Transportation Research Institute (UMTRI) and University of Michigan Logistics, Transportation & Parking Department, we redesign the University of Michigan Campus Transit System to adapt to the pandemic-influenced course schedules and student activities. We validate the results via simulation and further improve the design with practical considerations
● See more details in our paper [1]
● Visit our Online Bus Route Decision Tool [2]
○ Upload the distance between stops and number of buses
○ Download the optimized bus routes
Watch the video for an overview:
Watch the video for an overview:
Simulating aerosolized particles to improve bus safety during COVID-19
Simulating aerosolized particles to improve bus safety during COVID-19
By simulating how particles exhaled from passengers on a bus would travel through the vehicle under various conditions, researchers at the University of Michigan have been able to develop guidelines for how to operate campus buses more safely during the COVID-19 pandemic.
By simulating how particles exhaled from passengers on a bus would travel through the vehicle under various conditions, researchers at the University of Michigan have been able to develop guidelines for how to operate campus buses more safely during the COVID-19 pandemic.
Beginning Aug. 24, 2020, all bus riders and drivers must wear face coverings, windows will remain open, capacity will be reduced to nearly half and all previous routes will be replaced by a high-frequency hub-and-spoke routing system, among other changes.
Beginning Aug. 24, 2020, all bus riders and drivers must wear face coverings, windows will remain open, capacity will be reduced to nearly half and all previous routes will be replaced by a high-frequency hub-and-spoke routing system, among other changes.
The hub-and-spoke system relies on short, direct routes with fewer stops. The shorter routes not only limit rides to 15 minutes or less, they also enable more frequent service with the same number of buses, reducing wait times as well as the number of passengers on each bus.
The hub-and-spoke system relies on short, direct routes with fewer stops. The shorter routes not only limit rides to 15 minutes or less, they also enable more frequent service with the same number of buses, reducing wait times as well as the number of passengers on each bus.
Overview of the Network Redesign
Overview of the Network Redesign
Provided by social-distancing guidelines, each campus bus cannot hold more than 50% capacity (or even lower) and each passenger is desired to ride no more than 15 minutes in the same bus. One idea is to change the current system with long routes into a hub-and-spoke system [4] to (i) shorten routes that connect the 4 campuses of University of Michigan at Ann Arbor and (ii) increase overall bus capacity utilization and frequency. Furthermore, we consolidate bus stops to reduce the number of stops and shorten loading/unloading time.
Provided by social-distancing guidelines, each campus bus cannot hold more than 50% capacity (or even lower) and each passenger is desired to ride no more than 15 minutes in the same bus. One idea is to change the current system with long routes into a hub-and-spoke system [4] to (i) shorten routes that connect the 4 campuses of University of Michigan at Ann Arbor and (ii) increase overall bus capacity utilization and frequency. Furthermore, we consolidate bus stops to reduce the number of stops and shorten loading/unloading time.
We redesign the system by:
We redesign the system by:
- Estimating critical demand for different origin-destination (O-D) pairs and essential bus stops.
- Constructing a hub-and-spoke system and building an integer programming model to generate routes.
- Running simulation to evaluate the performance of the new system.
- Modifying the routes and stops according to practical restrictions.
1. Demand Estimation
1. Demand Estimation
By analyzing enrollment data, housing data, future class & activity schedules, parking space count data, and so on, we estimate critical demand for different O-D pairs and conclude them as trip patterns in various time over the day. We further map them to time series of demand at essential bus stops and identify the peak time.
By analyzing enrollment data, housing data, future class & activity schedules, parking space count data, and so on, we estimate critical demand for different O-D pairs and conclude them as trip patterns in various time over the day. We further map them to time series of demand at essential bus stops and identify the peak time.
Trip Patterns in the Morning
Trip Patterns at Noon
Trip Patterns in the Afternoon
2. Hub-and-spoke System Design
2. Hub-and-spoke System Design
We build an integer programming model to link necessary stops into routes that connect hubs (i.e., main bus stops across multiple campuses) and stops within individual campuses. (The solution reduces the number of bus stops in use by 50% and changes the original long routes into a hub-and-spoke system type.) In the model, we minimize the total travel time of all routes and the number of routes used to cover all stops, where a stop is covered if it is either included in any route or within 5-min walking to any included stop. We also add constraints to restrict that (i) each route visits its assigned stops one by one; (ii) each route returns to its hub after visiting all assigned stops; (iv) the trip time of any route is less than 15 minutes.
We build an integer programming model to link necessary stops into routes that connect hubs (i.e., main bus stops across multiple campuses) and stops within individual campuses. (The solution reduces the number of bus stops in use by 50% and changes the original long routes into a hub-and-spoke system type.) In the model, we minimize the total travel time of all routes and the number of routes used to cover all stops, where a stop is covered if it is either included in any route or within 5-min walking to any included stop. We also add constraints to restrict that (i) each route visits its assigned stops one by one; (ii) each route returns to its hub after visiting all assigned stops; (iv) the trip time of any route is less than 15 minutes.
Mathematical Formulation
Redesigned Transit System
In the redesigned system, we recommend six new routes (as shown on the right):
In the redesigned system, we recommend six new routes (as shown on the right):
- Main Commute: connecting the Central and the North campuses
- North 1, 2, 3: covering demand in the North campus
- Central 1, 2: covering demand in the Central campus
3. Simulation Study
3. Simulation Study
We use ProModel [3] for simulation studies.
We use ProModel [3] for simulation studies.
Input: stops and time schedules, hourly ride patterns and rates, travel time in between stops, and loading/unloading time.
Input: stops and time schedules, hourly ride patterns and rates, travel time in between stops, and loading/unloading time.
Output: comparison between the original and the recommended systems on: the number of served rides, real time busload track & bus utilization rate, the number of passengers waiting at each stop, and waiting and traveling time of passengers.
Output: comparison between the original and the recommended systems on: the number of served rides, real time busload track & bus utilization rate, the number of passengers waiting at each stop, and waiting and traveling time of passengers.
We further implement sensitivity analysis by varying bus capacity from 100% to 50% and then 25%, and varying bus availability/frequency. To simulate the performance when facing unexpected disruptions, we also perform stress tests under vulnerable events, e.g. lack of drivers, mechanical issues, and peak hour demand.
We further implement sensitivity analysis by varying bus capacity from 100% to 50% and then 25%, and varying bus availability/frequency. To simulate the performance when facing unexpected disruptions, we also perform stress tests under vulnerable events, e.g. lack of drivers, mechanical issues, and peak hour demand.
Example: Original Routes Commuter North & South VS Recommended Routes North 2, Main Commute, and Central 2
Example: Original Routes Commuter North & South VS Recommended Routes North 2, Main Commute, and Central 2
Commute North & Commuter South
North 2, Main Commute, and Central 2
The extensive simulation results indicate that the hub-and-spoke system outperforms the original system because:
The extensive simulation results indicate that the hub-and-spoke system outperforms the original system because:
- It allows more frequent buses assigned to each route as the longest drive for any trip is reduced to 11 to 14 minutes, as compared to the 45 or 55 minutes in the original system.
- The total riding capacity is significantly increased even the number of seats we can use on each bus is reduced to 25% of the full capacity.
- It can satisfy the original peak demand (assuming all in-person classes in the Fall semester) using half or even lower bus capacity and the same number of buses.
- On average, 35% of passengers wait shorter than 1 minute, 70% wait less than 5 minutes, and 70% riders reach their destinations without transit.
- > 99% of passengers can complete their trips without being on any bus for longer than 15 minutes, which satisfies the guideline for reducing infection risk.
4. Practical Adoption
4. Practical Adoption
The routing solution generated by the mathematical model are further modified and improved via
The routing solution generated by the mathematical model are further modified and improved via
- Simulation results
- Professional advice from experts considering transportation infrastructure and driving experiences
Example: North 1
Example: North 1
Generated by Model
Modified by Simulation and Experts
Conclusion and Next Step
Conclusion and Next Step
This work inspires us to re-consider and re-design existing transportation systems and their operations to mitigate virus spread in cities and communities. The new systems need to satisfy diverse mobility demand under varying supply while meeting the reduced capacities due to lockdown and social distancing. It is of vital importance to learn and improve operational efficiency of these new systems, in the face of unexpected disruptions both during and post the pandemic.
This work inspires us to re-consider and re-design existing transportation systems and their operations to mitigate virus spread in cities and communities. The new systems need to satisfy diverse mobility demand under varying supply while meeting the reduced capacities due to lockdown and social distancing. It is of vital importance to learn and improve operational efficiency of these new systems, in the face of unexpected disruptions both during and post the pandemic.
References
References
[1] Chen, G., Fei, X., Jia, H., Yu, X., Shen, S., "An Optimization-and-Simulation Framework for Redesigning University Campus Bus System with Social Distancing and Uncertain Demand," arXiv preprint arXiv:2010.10630 (2019).
[1] Chen, G., Fei, X., Jia, H., Yu, X., Shen, S., "An Optimization-and-Simulation Framework for Redesigning University Campus Bus System with Social Distancing and Uncertain Demand," arXiv preprint arXiv:2010.10630 (2019).
[2] An Optimization Model for Finding Optimal Bus Routes. http://covid19reopening.northcentralus.cloudapp.azure.com:5000/busroute
[2] An Optimization Model for Finding Optimal Bus Routes. http://covid19reopening.northcentralus.cloudapp.azure.com:5000/busroute
[3] Bryan, D. L., & O'kelly, M. E. (1999). Hub‐and‐spoke networks in air transportation: an analytical review. Journal of regional science, 39(2), 275-295.
[3] Bryan, D. L., & O'kelly, M. E. (1999). Hub‐and‐spoke networks in air transportation: an analytical review. Journal of regional science, 39(2), 275-295.
Student Contributors:
Student Contributors:
Gongyu Chen
Gongyu Chen
Undergraduate, Department of Industrial and Operations Engineering
Undergraduate, Department of Industrial and Operations Engineering
University of Michigan
University of Michigan
Xinyu Fei
Xinyu Fei
PhD, Department of Industrial and Operations Engineering
PhD, Department of Industrial and Operations Engineering
University of Michigan
University of Michigan
Huiwen Jia
Huiwen Jia
PhD, Department of Industrial and Operations Engineering
PhD, Department of Industrial and Operations Engineering
University of Michigan
University of Michigan
Xian Yu
Xian Yu
PhD, Department of Industrial and Operations Engineering
PhD, Department of Industrial and Operations Engineering
University of Michigan
University of Michigan
Page updated
Report abuse