The National Institutes of Health (NIH) is located on what is known as the NIH campus. This campus is huge and consists of many buildings, including 30 parking garages. Over 12,000 vehicles enter this campus on an average day, meaning that optimizing traffic flow and planning well-designed parking garages are essential to ensuring that this is an effectively run center.
Problem
NIH was interested in two parking garages on the south side of the campus. The first is a four-story parking garage, however, there is no vehicle movement between the floors, and each floor has unique entrance and exit points. It is simply four parking lots on top of each other and is termed MLP7.
The second, which is a surface lot with 241 spaces, is going to be replaced by a six-story parking garage with 1,420 spaces and will be known as MLP12.
NIH wanted to understand the impact of different entrance combinations for MLP12 and whether it would have sufficient parking when completed. They also wanted to focus on optimizing traffic flow around both MLP7 and MLP12.
Key metrics that NIH was interested in included the length of time for a vehicle to park and the length of time to leave after parking. Additionally, the number of stops and total time stopped for each vehicle needed to be considered.
Covid-19 caused some data challenges, and so data from a suitable period needed to be found to run pre-pandemic scenarios to prepare for the easing of restrictions.
Solution
Mosimtec, a company that offers consulting and simulation modeling services to businesses around the world, developed a simulation model to answer the required questions. There were four key phases that the developers worked through in the project:
- Functional specification – both Mosimtec and NIH worked together on the project scope and objectives.
- Dynamic simulation model – a custom-developed AnyLogic model was created with the necessary system components, logic, and behavior.
- Scenario analysis – an analysis was conducted of pre-defined scenarios identified in the first phase.
- Training and knowledge transfer – Mosimtec taught NIH how to use the model and run scenarios so that they could continue to use the model in the future.
The model uses Excel for both inputs and outputs. Inputs include probabilistic travel routes, parking destinations, parking times, and others. All scenarios are loaded into the model from Excel, and after model replications are complete, the results are exported back to Excel for analysis and reporting.
The developers created a two-dimensional space to allow simultaneous viewing of all floors from an animation perspective instead of having one road network in 3D. MLP12 is still in the design phase, so there can be changes to the plan of this parking garage, such as the exact entrance location on any floor. By implementing a road network design for each floor, these changes can be made easily without affecting other sections of the model.
Another reason for having a road network for each floor of the parking garages is for model logic and output purposes. It is easier to have real-time statistics and insights for each floor by using separate road network designs for each one.
In the model, vehicle movement is categorized into four main areas:
- Creating vehicles.
- Routing to a parking garage.
- Parking in an available space.
- Exiting the system.
An important consideration was how to control the flow of vehicles, particularly at intersections. This was achieved by using the Road Traffic Library and employing traffic lights to represent stop signs, resulting in alternating flows. This library was also used to implement the multiple road network designs.
Transforming the multi-story parking garages into a road network in two-dimensional space required the use of multiple HashMaps. On model initialization, these were populated from the Excel inputs, resulting in an improved model run speed. Also, using HashMaps like this meant that the Excel front-end can be changed easily if there are any alterations to the parking garage architecture. As a result, there won’t have to be any changes to the code in the AnyLogic model.
There were over ten different HashMaps used, but the most essential for optimizing the flow of traffic are illustrated in the table below.
Mosimtec worked with NIH to design the Excel front end, which included all the inputs, output KPIs, and dashboards. Working together ensured that NIH would be able to understand and use the model effectively. Some of the inputs and outputs of the model are illustrated below.
When running the model, it is possible to observe the vehicles traveling to the entrances of the parking garages as well as the movement within these while they are parking and then leaving the garages and the system.
It is important to remember that in this model, MLP7 doesn’t allow movement between different floors. In MLP12, the vehicles move up and down and across the floors.
The different floors of the parking garages are displayed side by side for ease of viewing. The floor on the main network is the top floor, cascading down from left to right, as illustrated in the second picture below for MLP12.
Each road network design uses a density map, which displays different colors depending on the amount of traffic. This helps users visually understand what is happening in the model before analyzing the results.
The logic is also displayed. There is a road descriptor for each network: creation, routing, MLP12 movement and parking, and finally MLP7 parking and other lots, as well as the vehicles exiting the system.
Results
AnyLogic, being used as transportation planning software, allowed NIH to understand that, in the pre-Covid-19 state, the proposed combinations for MLP12 did not affect the traffic on surrounding roads significantly.
NIH also compared the different metrics of the vehicles through the system. These were considered pivotal because NIH wanted to continue prioritizing the on-campus experience for their employees and visitors.
NIH has plans to continue exploring different entrance combinations for the planned design of the parking garage MLP12 and optimizing the traffic flow in the southern part of the campus. Any changes that they decide to make can be seamlessly adopted and implemented because of how the model was designed and developed. Additionally, as Covid-19 restrictions ease, the number of vehicles, waiting times, and duration of time vehicles are parked can also be updated with new data.
Finally, the model could be expanded from the south of campus to the whole campus and even further to surrounding roads.
The case study was presented by Geoff Skipton and Yusuke Legard, of Mosimtec, at the AnyLogic Conference 2022.
The slides are available as a PDF.
