AWS HPC Blog
Harnessing the power of agent-based models for mitigating supply chain risks and managing costs
Supply chains today face major vulnerabilities due to their reliance on traditional methods that only use historical data. This can leave organizations exposed to disruptions or unable to effectively identify weaknesses and prepare for future shocks. The increasing complexity of global supply networks makes this challenge even more acute.
Imagine having complete visibility into your supply chain dynamics, with the ability to stress test against any possible disruption scenario. Picture being able to simultaneously optimize for efficiency, speed, and sustainability while building resilience. The ideal state is a supply chain that can both withstand shocks and perform excellently under normal conditions.
Today we’ll discuss a new simulation-based approach using agent-based modeling (ABM), rich data sets, and advanced analytics that makes this possible. Through a collaboration between AWS, Simudyne, and Moody’s, organizations can now model their current supply networks in detail and run sophisticated simulations of potential disruptions. This allows them to identify vulnerabilities before they become problems and implement targeted mitigation strategies. The solution enables organizations to optimize performance across multiple metrics while making data-driven decisions about supply chain management. This solution combines AWS’s cloud infrastructure, Simudyne’s ABM expertise, and Moody’s risk analytics to deliver insights into supply chain dynamics, enabling both resilience and efficiency improvements.
Understanding agent-based models in the context of supply chains
Traditional simulation approaches like discrete event simulation and system dynamics struggle with modern supply chain complexity. Agent-based modeling (ABM) provides a powerful alternative paradigm to model heterogeneous behaviors and intricate interdependencies.
An ABM consists of autonomous “agents” representing businesses, intermediaries, consumers, etc., operating in a shared environment. Agents follow rules to make decisions and interact with each other and their surroundings. ABMs excel at simulating emergent phenomena arising from multi-agent interactions at a micro-level. This bottom-up approach captures non-linear dynamics, feedback loops, and tipping points often missed by top-down abstractions.
As supply chains become globalized, distributed, and impacted by interconnected risks, heterogeneity, and entropy increase. In uncertain, complex environments, ABMs provide an intuitive and efficient technique to explore “what-if” scenarios. By explicitly modeling granular agent decisions and interactions, ABMs reveal how localized disruptions cascade through networks. This granularity in simulating multi-agent adaptive behaviors is pivotal for enhancing supply chain resilience and robustness. Against emerging risks, ABMs enable understanding systemic vulnerabilities through a bottom-up lens on interacting autonomous agents.
The unique advantages of ABMs for supply chain analysis
The strengths of ABMs in supply chain analysis hinge on their ability to encapsulate the complexity and dynamism of supply chain systems. Unlike traditional frameworks like system dynamics models which usually make simplifying assumptions about interactions, ABMs thrive on complexity and interdependencies. They uniquely facilitate the exploration of emergent behaviors — unexpected outcomes resulting from the interaction of agents within the system. This characteristic is especially valuable in understanding how minor disruptions can cascade through a supply chain, potentially unveiling vulnerabilities and inefficiencies not easily captured under more aggregate analysis methods.
ABMs also excel in scenario planning, allowing supply chain managers to rigorously evaluate the impact across three major categories: 1) Descriptive or operational scenarios for normal conditions, 2) Disruptive scenarios like natural disasters or geopolitical events, and 3) Strategic decision scenarios around factors they can control. This enables assessing supply chain resilience in a controlled, virtual environment under various disaster scenarios ranging from natural calamities to geopolitical tensions. By enabling a detailed examination of the ripple effects from such disruptive events, as well as modeling strategic decisions, ABMs empower decision-makers with the insights needed to focus their efforts and fortify supply chains against future disruptions while optimizing performance under normal operations.
Moreover, the adaptability of ABMs to incorporate real-time data stands as a significant advantage, ensuring analyses remain timely and relevant and actionable in the face of rapidly changing market conditions. This agility is paramount in today’s volatile economic landscape, underscored by Moody’s ability to help our clients identify and measure the increasing frequency and severity of supply chain disruptions. In essence, ABMs offer a forward-looking, dynamic lens through which supply chains can be examined, analyzed, and optimized for resilience and efficiency.
- Impact of disruptions: By modeling disruptive events like cyber-attacks or natural disasters, ABM provides a detailed understanding of their potential magnitude and ripple effects, aiding in effective contingency planning.
- Behavioral interactions: ABM captures the interactions between different agents, revealing hidden dependencies and potential points of failure that might not be apparent through traditional analysis methods.
- Optimization of response strategies: ABM allows for the simulation and evaluation of various response strategies, identifying the most effective approaches for quick recovery and minimizing downtime.
- Predictive analytics: Using historical data and predictive algorithms, ABM can forecast potential future disruptions and their likely impacts, enabling better preparation and resource allocation.
- Supply Chain resilience: By testing the system’s ability to withstand and recover from disruptions, ABM helps identify and strengthen weak links within the supply chain, leading to a more robust and resilient operation.
- Cost-Benefit analysis: ABM can perform cost-benefit analyses of different risk mitigation strategies, ensuring economically viable decisions and optimal resource allocation.
Case Studies: ABMs in action – oil and gas simulation
Moody’s participated in an AWS Digital Innovation engagement to apply the “working backwards” methodology to define an oil and gas supply chain disruption ABM. The engagement involved business owners and product teams. They first evaluated user needs, using the Colonial Pipeline case as a core example. The teams developed a solution based on prioritized scenarios leveraging Simudyne Supply Chain Toolkit available on AWS Marketplace. Fig 1.
In the case of the Colonial Pipeline, an ABM can explicitly represent the different stakeholders involved – the pipeline operator, refineries, storage terminals, retailers, transportation providers, and end consumers. The modeled agents make operational decisions based on their local information, objectives, and constraints. This allows exploring how disruptions like pipeline shutdowns can cascade through the system in unintuitive ways as companies scramble to find alternative supply routes. Critically, an ABM can incorporate each party’s adaptive responses and mitigation strategies like shifting to truck/rail delivery.
By allowing users to visualize and compare a range of “what-if” scenarios, agent-based models provide unique insights for supply chain risk management. They support the understanding of systemic vulnerabilities, stress testing contingency plans, and evaluating the costs/benefits of potential mitigations before disruptions occur. Optimizing strategies identified through ABM simulations has the potential to improve supply chain efficiency by 15-25%, reduce response times to disruptions by 30-50%, and enhance long-term sustainability metrics like emissions/waste by 10-20%.
This analytical capability, combining rich scenario modeling with quantifiable performance gains, represents a powerful decision tool for increasing the resilience and adaptive capacity of petroleum, manufacturing, and other strategic supply chains. The ability to rigorously assess tradeoffs between mitigation costs and operational improvements is invaluable.

Fig 1: As a first step in defining the supply chain simulation service for Oil and Gas, Moody’s and Simudyne participated in an AWS Digital Innovation engagement to apply the “working backwards” methodology. The engagement involved industry experts, business owners and product teams. The teams first evaluated user needs, using the Colonial Pipeline case as a core example. They developed a solution based on prioritized scenarios (see screenshot below) leveraging Simudyne Supply Chain Toolkit.
AWS reference architecture: leveraging Simudyne supply chain toolkit
A deployed solution utilizes an Amazon Simple Queue Service (SQS), enabling the queuing of many ABM simulation requests. One of the key advantages of this architecture is the added layer of security provided by the client sending messages directly to the SQS. This approach eliminates the need for external internet connections to the data and models, effectively safeguarding sensitive information.
Amazon EventBridge actively monitors the SQS queue, and when a new message is detected, an Elastic Container Service (ECS) task is triggered. This task dynamically provisions an Elastic Compute Cloud (EC2) instance within a private subnet, ensuring that it remains inaccessible from the internet.
The ECS task then loads the Simudyne SDK Amazon Machine Image (AMI) available through the AWS Marketplace, as well as the supply chain ABM model and any associated customer data from an Amazon Simple Storage Service (S3) bucket. All of these components are meticulously separated and hidden from external access, further enhancing the overall security posture.
Once the ABM simulation is complete, the resulting data is securely stored in a separate S3 bucket accessible from an ECS service EC2 instance. This service provides a reliable backend server that can seamlessly restart or move between subnets in the event of system crashes or power outages, achieving uninterrupted operation. The backend server is responsible for post-processing the ABM results, generating valuable analytics and insights that users can conveniently access through their client applications.
By leveraging the scalability, security, and reliability of AWS services, the Simudyne SDK offers a robust and efficient platform for conducting complex supply chain simulations, empowering businesses with valuable insights to drive informed decision-making.

Fig 2: A deployed solution utilizes Amazon SQS for queuing ABM simulation requests, with clients sending messages directly to SQS, adding a security layer by eliminating the need for external internet connections to sensitive data and models. Amazon EventBridge monitors the SQS queue and triggers an ECS task to provision an EC2 instance within a private subnet to run the Simudyne SDK AMI from AWS Marketplace, loading the supply chain ABM model and data from a secure S3 bucket, with results stored in a separate S3 bucket accessed by a reliable ECS service EC2 backend server.
Integrating ABMs with traditional supply chain management practices
Combining the dynamic simulations of agent-based models with the solid data-driven foundation of traditional supply chain management techniques offers an innovative approach to navigate the unpredictable events of today’s markets. The historical data and quantitative analysis traditionally used lay a reliable groundwork for grasping past trends and consistent operations. However, these methods often struggle to foresee the intricate dynamics and spontaneous patterns prevalent in today’s supply chains. ABMs step in to fill this gap by enabling a simulation of individual agents’ actions and their interplays, enhancing the understanding of complex system behaviors.
For a successful integration of ABMs and conventional strategies, organizations are recommended to embrace a blend of methodologies. This blend should capitalize on the detailed scenarios and potential future stressors identified through ABMs, augmented by the data-rich insights from traditional supply and demand analytics for everyday decision-making and performance tracking.
Steps to implementing an agent-based model in the context of supply chain disruptions
Embarking on the journey of implementing an agent-based model for your supply chain involves a structured approach to fully leverage its potential. Initially, identify the specific objectives you aim to achieve with ABM, such as enhancing resilience, improving efficiency, or identifying vulnerabilities. Next, gather detailed data on your supply chain operations, including supplier networks, logistics, production processes, and demand forecasts. This data forms the foundation of your model, enabling a realistic simulation of your supply chain dynamics.
- Define the objective: clarify what you aim to achieve with your simulation. In this case, you want to assess how various supply chain disruptions (due to weather events, policy changes, or economic fluctuations) impact your operations.
- Understand key factors: identify the key factors that influence your supply chain performance. This may include operational metrics, economic conditions, and industry-specific factors.
- Design your agent-based model: an agent-based model consists of autonomous agents, each with its own set of characteristics and behaviors, operating within a defined environment. Agents: In a supply chain context, agents can represent companies, regulatory bodies, consumers, etc. Define the attributes (e.g., size, location) and behaviors (e.g., decision-making processes, responses to disruptions) of each type of agent.
- Environment: the environment includes everything external to the agents, such as weather conditions, policy landscapes, and economic contexts. It should allow for dynamic changes affecting agents. Interactions: Define how agents interact with each other and the environment. This includes transactions, contracts, and responses to external shocks.
- Incorporate performance assessment Module: develop a module within your simulation that evaluates the performance of the agents involved in the supply chain. This should consider the identified key factors and translate them into impacts on agents’ operations, decision-making, or outcomes.
- Develop the simulation: using a programming language suited for ABMs (e.g., Simudyne Supply Chain Toolkit from the AWS Marketplace), develop your simulation. This involves coding the agents, their environment, interactions, and the integration of the performance assessment module.
- Validate and calibrate your model: compare simulation outcomes with historical data to ensure your model accurately reflects real-world dynamics. Adjust as necessary. Calibration: Fine-tune your model parameters so that the outcomes align with expected behaviors under known conditions.
- Experimentation and analysis: conduct experiments by varying input parameters, such as the severity of weather events or the degree of policy change, to understand their impacts on your supply chain operations. Analyze how changes in key factors influence outcomes.
- Reporting and iteration: document your findings and iteratively refine your simulation based on feedback and new data. This might involve incorporating more detailed factor data or adjusting agent behaviors to more closely mirror real-world observations.
By systematically developing an ABM with a performance assessment module at its core, customers can achieve nuanced insights into how supply chain disruptions impact operations. This process requires a deep understanding of both the supply chain dynamics at play and the comprehensive factors that impact performance.
The future of supply chain resilience: the role of ABMs
As we navigate into the future, the role of agent-based models in enhancing supply chain resilience and efficiency is poised to become increasingly pivotal. The adaptive nature of ABMs, with their ability to simulate complex, evolving scenarios involving myriad interacting agents, offers a forward-looking lens through which businesses can preemptively address potential supply chain deficiencies. In an environment where the pace of change is accelerating, driven by technological advancements, environmental concerns, and geopolitical shifts, the agility afforded by ABMs will be indispensable. Moreover, the integration of ABMs with emerging technologies such as artificial intelligence (AI) and machine learning (ML) heralds a new era of predictive analytics, enabling supply chains to not only respond to disruptions but to anticipate them with unprecedented precision. As businesses strive to remain competitive in a landscape characterized by uncertainty, the strategic implementation of ABMs will be crucial for fostering robust, resilient supply chains capable of withstanding the challenges of tomorrow. Through continuous innovation and adaptation, ABMs stand at the forefront of evolving supply chain strategies, heralding a future where resilience is not just reactive, but inherently built into the fabric of supply chain management.
“In this era of interconnected risk corporations want to understand the impacts of a much wider set of risks to their supply chains, and the compounding effect of these risks on one another. We think that the combination of agent-based model simulation and Moody’s data can be an important tool to help corporations navigate the impact of interconnected risks.”
Masha Muzyka, Global Head of Industry Practice, Moody’s