Uncovering the differences in Agent Based and Discrete Event Modelling
Agent Based Models (ABMs)
Agent Based Models (ABMs) simulate the behaviour of individual agents -such as machines, workers, or vehicles -within a defined environment. Each agent follows a unique set of rules that dictate how it interacts with other agents and its surroundings. Through these interactions, complex patterns emerge at the system level, revealing insights that are not obvious when analysing individual components in isolation.
The Value of Agent Based models
ABMs excel at capturing the dynamics of local decision-making and the variability of individual agents operating in a complex environment. This approach is particularly valuable for adaptive systems where the actions of individual agents can propagate non-linearly and lead to significant, system-wide outcomes.
In dynamic environments like underground mining, ABMs are particularly useful for modelling fluctuating conditions and external factors, leading to more accurate predictions and enabling adaptive decisions.
For instance:
ABMs allow organisations to analyse how small changes in agent behaviour -like a truck’s speed adjustment due to road congestion- impact overall productivity.
By accounting for individual variability (e.g., minute-to-minute machine operational conditions, operator behaviours), ABMs provide more precise predictions compared to traditional models, that may assume more uniform behaviours across agents.
Use in Mining
Consider simulating haul truck movements within an underground mining operation:
Each haul truck acts as an independent agent, making real-time decisions based on conditions such as road congestion, task priority, or equipment availability.
Trucks interact with other agents, like boggers or jumbos, and adjust their paths or speeds depending on proximity to other equipment, the number of agents on the route, or the location of obstacles.
This dynamic setup mirrors the unpredictability of real-world conditions, where each truck responds uniquely in response to local conditions, producing emerging patterns on system performance.
Typical Inputs and Outputs in Mining Agent-Based Models:
Inputs:
Road Network: Mapping of travel routes and bottlenecks in the mining area, location of stockpiles, and areas of activity.
Equipment Specifications: Details of trucks, boggers, and jumbos, including their operational parameters.
Load Distributions: Data on the individual-level distributions of load sizes, and loading times.
Agent Rules: Rules governing each agent’s decision-making, such as task prioritisation and assignments and right of way rules.
Outputs:
Cycle Times: The time taken for each individual truck to complete its loading and hauling cycle, offering insights into operational efficiency.
Material Throughput: The quantity of material moved over time, a key metric for productivity analysis.
Congestion Heatmaps: Visualisation of high-traffic areas, helping to identify bottlenecks and optimise routing and scheduling.
ABMs, by leveraging individual agents and rule-based interactions, provide mining operations with powerful insights into complex, real-time scenarios. By integrating these models into platforms like DiiMOS™, companies can enhance operational efficiency, reduce bottlenecks, and achieve optimised workflows that adapt in response to the variable conditions of modern industrial environments in a way that is not achievable by other modelling approaches.
Discrete Event Models (DEMs)
Discrete Event Models (DEMs) are simulations that focus on the timing, sequence, and interdependencies of events in a system. In DEMs, the state of the system only changes at specific moments when key events occur.
For instance, in a mining operation, events could include "material arriving at a crusher," "crush complete," or "conveyor starts." These models operate in a process-driven framework, meaning they track sequences where tasks happen in a relatively set order with dependencies. DEMs are ideal for environments where the process flow and timing of operations are structured and predictable.
The Value of Discrete Event Models
DEMs highly effective for optimising workflows and increasing efficiency in systems with well-defined, sequenced tasks. This modelling approach is valuable because it allows for:
Workflow Optimisation: DEMs identify the sequence of tasks, known as the critical path, that must be completed on time for the system to run smoothly and meet targets within expected timeframes. By analysing timing, DEMs help highlight bottlenecks and areas for process improvement.
Efficiency in Process-Driven Operations: DEMs provide insights into system-wide efficiency by showing how delays in certain events affect the overall flow. This capability is useful for minimizing wait times, optimising task sequencing, and identifying dependencies within workflows.
Identification of Critical Paths: DEMs trace the steps in a sequence that are most time-sensitive, allowing for targeted adjustments that minimise downtime and ensure continuous operations.
Use in Mining
A typical application of DEMs in mining would be to simulate the flow of material from a crusher to a conveyor system:
Events in this case could include the arrival of material at the crusher, completion of crushing, material transfer to the conveyor, and the conveyor starting operation.
The DEM tracks these events in sequence, following a predetermined workflow. This approach highlights the efficiency of material processing but does not capture fine-grained variations in equipment behaviour, such as the individual variability in how each crusher performs.
By analysing event sequences, DEMs can reveal delays in processes (like prolonged downtime for a crusher) that could hinder throughput and suggest improvements to scheduling and maintenance to reduce these impacts.
Typical Inputs and Outputs in Mining Discrete Event Models:
Inputs:
Sequence of Planned Events: List of tasks in the process (e.g., material loading, crushing, conveying) and the order in which they occur.
Event Dependencies: Relationships between events that determine when each task can start (e.g., conveyor cannot begin until the crusher has completed its task) and end.
Probabilities of Task Durations: Expected times for each event, often based on historical data, which can include variability for cycle times.
Outputs:
Material Throughput: The volume of material processed over time, useful for understanding system productivity.
Scheduled Downtime: Time allocated for maintenance or other planned interruptions, helping to measure and reduce idle times.
Critical Path: The sequence of events most crucial to keeping operations on schedule, identifying where delays would have the greatest impact on overall system performance.
Discrete Event Models, with their focus on structured, process-driven workflows, offer a strategic approach to improving task efficiency and resource allocation in mining and other industrial settings. When integrated into platforms like DiiMOS™, DEMs provide a clear pathway to streamlined operations, reduced downtime, and optimized material flow in complex, interdependent systems.
Spotting the Differences in Action
“Does the model track individual agents (like trucks or boggers) and their unique behaviours, or does it only schedule tasks based on fixed events?"
A true Agent-Based Model will focus on the behaviour of individual entities, while a discrete event model will focus on timing and sequences of larger scale processes.
"How does the model handle variability or unexpected events? For example, how does it adapt if one truck breaks down or if road conditions change?"
Agent-Based Models will dynamically adjust based on the behaviour of individual agents, while Discrete Event Models may struggle with unexpected variability.
“Can the model simulate local decision-making and interactions between machines, or does the model follow a predefined sequence?"
If the system is driven by local decision-making (i.e., agents can choose their actions based on their surroundings), it’s likely an Agent-Based Model, whereas predefined sequences point to a Discrete Event Model.
Using the right model for the job
The choice between Agent-Based Models and Discrete Event Models can significantly influence productivity, adaptability, and risk management. Choosing the right model determines how well your simulation can replicate real-world complexities and respond to variability. An Agent-Based Model offers much greater flexibility and insight in dynamic environments where individual behaviours and local interactions drive outcomes, enabling more adaptive and responsive decision-making. In contrast, a Discrete Event Model provides clarity and efficiency for structured, process-driven workflows, optimising resource allocation and minimising delays in predictable systems. Selecting the appropriate approach ensures that your operation can achieve maximum efficiency while effectively managing risks inherent in your environment
Agent Based Models
ABMs are essential when variability and near real-time decision-making drive performance. For example, when equipment like trucks and boggers must coordinate in response to road blockages or breakdowns, ABMs model these interactions to optimise efficiency.
Key Question: "How do you handle unpredictability in your operation?”
Agent-Based Models provide a way to simulate individual behaviours under varying conditions, enabling better resource allocation and management in highly dynamic operations.
Discrete Event Models
Discrete Event Models are better suited for highly predictable processes, such as material handling and production workflows. If your goal is process optimisation, DEMs help reduce bottlenecks by scheduling tasks precisely.
Key Question: How do I optimise a process?
For structured workflows, Discrete Event Models deliver efficiency by streamlining task sequences and managing dependencies.
A place for both
Agent-Based Models excel at capturing complex, adaptive behaviours, while Discrete Event Models are highly effective for structured, sequential operations. Both modelling approaches have unique strengths for different types of analyses within mining, and can potentially be combined to leverage each other’s capabilities.
DiiMOS™ is an integrated decision optimisation platform designed to solve complex problems in your operations.
DiiMOS™ lets you model your operation for enhanced visibility, navigate with predictive insights, simulate scenarios for effective planning, and optimise processes to make the best decisions.
Our solution arms you with a powerful operations planning and forecasting capability, with modular design to grow with your needs along the mining value chain, from ore-body to customer.
With our approach, you’re not just choosing a tool; you’re choosing a future where planning decisions are faster, smarter, and better integrated, unlocking value drivers related to cost, productivity, emissions, and energy directly to the bottom line.