Simulation for Transportation Network Optimization via Truck Yard Revision

Simulation for Transportation Network Optimization via Truck Yard Revision

Amazon is a common platform for online shopping in the USA and Europe. Many purchases and deliveries go as smoothly as they are supposed to, but some may be delayed. The company does its best to minimize such delays and other customer inconveniences. It works on transportation network optimization, looks for new solutions for supply chain risk mitigation, and implements the best delivery scenarios.

Nevertheless, there is still room for improvement, especially during high periods (Prime Day Sale, Black Friday, Christmas, etc.). The company focuses on making sure that orders arrive on time and without any issues, regardless of the number they receive.

Problem

Some delays are caused by issues in the fulfillment center truck yards. These are the places where trucks and trailers come and go. Yards are filled with constant movements, people, and goods.

The issue is chaotic behavior. The situation may be compared to airport delays. For example, passengers are in a plane waiting for take-off, but suddenly, the pilot says there's a weather problem, and they are forced to stay in the taxiway for a couple of hours. The number of aircraft increases, which leads to a queue. As a result, there are multiple delays at various destination airports. The same happens in Amazon yards, and the analytical team works to resolve it.

A diagram of Amazon transportation network devided into first mile, middle mile, and last mile stage

Amazon transportation network breakdown by workstream

Amazon's transportation network has three stages: the first mile, the middle mile, and the last mile.

In the first mile, Amazon receives goods from suppliers and sends them to the middle mile stage. In the middle mile, goods are transported. Once the customer clicks “Order” on Amazon, that unit from the fulfillment center becomes available, gets packed and sent to a sorting center, and then it goes to a delivery station. After the station, the goods move to the last mile, reaching the customer.

The main challenge for transportation network optimization lies in the middle mile.

Amazon supply chain diagram with breakdown by scheduling

Amazon supply chain network breakdown by scheduling overview
(the golden square indicates the core process of this case study)

In the current Amazon supply chain department, there's a team that works on transportation network optimization; it schedules departures and ETA arrivals. Typically, 53-foot trailers transport goods many times. This helps mitigate supply chain risks and deliver goods on time. However, issues may occur due to the lack of truck yard optimization.

Amazon transportation facility

Amazon transportation (truck) yard scheme

When a trailer arrives at the yard, a transportation associate confirms the information at the guard shack. If valid, entry is granted. The driver either parks in the dock parking lot or uses an external yard for a temporary trailer stay. At the end of the day, the trailer is either emptied or loaded for departure. The transportation associate validates the information at the guard shack, and the truck leaves.

Stages of trailers movements through the Amazon truck yard

Trailer journey through the Amazon transportation yard

However, the yard has different aspects that may cause operational issues in the transportation network system, especially during peak seasons when customer demand increases. Sometimes, Amazon offices know about changes in advance, and other times, it's a last-minute surprise. So, the management needs a quick solution to handle these changes without messing up the transportation plans.

Solution

Schedules diaphragm showing dependancy of the gate queue and time

Example of four different schedules

Amazon's analytical team decided to use simulation modeling in AnyLogic to measure the transportation yard risks. So, they could test the hypothesis before breaking into current transportation network optimization.

Transportation yard schedules model

At first, analysts created a model to visualize schedule variations and see how those changes affect transportation yard operations.

Scenario 1 (Sc1) is smooth for the site, while Sc2, Sc3, and Sc4 cause a buildup of trucks at the gate, impacting traffic flow. This became a core metric for the team, and they tracked it closely to minimize external queue buildup. Sc1, Sc2, Sc3, and Sc4 are scenarios based on volume, headcount, and equipment availability, providing a metric for the customer.

Check-in process in the transportation yard model

A part of the transportation yard model with the truck check-in process

The model represents Amazon yard transportation processes. It guides trucks through different actions based on specific scenarios. AnyLogic models can operate multiple schedules at once, though in this case the team limited them to 140 schedules per experiment. Several iterations of schedules in the parameter variation experiment improved truck yard model operation, making transportation network optimization easier overall.

Traffic flow model

An aerial view of an Amazon truck yard

Traffic flow simulation

The team also created a traffic flow model using the AnyLogic Material Handling Library that has elements to model transporters and a storage system. This model represents a simplified version of the yard, with green items indicating trucks. Inbound loads are at the top, currently being unloaded, while outbound docks are being loaded at the bottom.

A safety feature was added to the model. It enables a red circle when a transporter spots a truck that is parking and blocking traffic.

The output of the traffic flow model represents the number of stopped traffic instances and the average speed of transporters within the yard. Safety is a top priority for Amazon, and headquarters wanted to know the overall speed in the yard, considering maximum speed, volume, headcount, and equipment availability at specific schedules and times.


Safety red circle that indicates a truck blocking traffic

Results

Transportation scheduling scenario validation

Metrics of scenarios validation in transportation scheduling model

Scenarios validation metrics (click to enlarge)

The table represents different scenario outcomes. With the AnyLogic model of truck yard schedules, the analytical team came up with a simple technical solution. They exported the model, so all Amazon offices could use it for parameter variation experiments. With validated inputs and data for building the model, they can get the results in one click.

The use of experiments results in quick improvements to the network processes. For example, if a Los Angeles branch faces issues with a new schedule and wants to assess the capacity and capability of the site, their analytical team runs scenarios in that model exported from AnyLogic. The structured data output helps them avoid potential risks and see why some approaches may not be used.

Such model export as a standalone Java application is available in AnyLogic Professional. Alternatively, AnyLogic Private Cloud enables running models online with complete control of access limitations.

Traffic flow model

With statistics on how all areas within the yard are used in different schedules, the analytical team created a model of the Amazon truck yard traffic flow.

Dependency of truck check-in queue and time. A graph showing hourly utilization breakdown

Click to enlarge charts

Based on the model, the team highlighted several conclusions:

  1. High dock door utilization affects gate queues: The dock door utilization consistently approaches the maximum threshold of 95%. As it gets closer to gridlock, the gate queue increases significantly after 48 hours of running the schedule.
  2. Parking slips are piling up: Even though the dock utilization seems fine, there's a buildup of parking slips. This increase in trucks is due to the dynamics of dock utilization. Many trucks need to enter the yard, but they can't due to the queue, leading to a long queue in front of the guard shack.

Instead of testing possible scenarios for transportation network optimization in real life, Amazon modeled them in AnyLogic simulation software. So, they saved time and funds, avoiding the cost of impractical solutions.

The case study was presented by Jae Yong Lee, of Amazon Transportation Services, at the AnyLogic Conference 2023. If you are interested in other Amazon cases, check out the Simulation-Driven Solution for Fulfillment Logistics Evaluation case study. It is about the impact of simulation modeling and AI integration on fulfillment logistics transformation.

The slides are available as a PDF.


Similar case studies

More case studies

Get a brochure with industry case studies

Download