IEE 475: Lecture A2 (2026-08-27): Introduction to Simulation Modeling

Overview

IEE 475 introduces simulation modeling by distinguishing models that answer what-if questions, static outcomes from dynamic trajectories, and three major dynamic-modeling approaches: system dynamics, agent-based modeling, and discrete-event simulation. The lecture focuses on discrete-event simulation, explaining how entities, resources, state, events, activities, and delays fit together, and why stochastic inputs can approximate complex real-world variability.

Key takeaways

A static model produces an outcome or outcome distribution, whereas a dynamic model produces a time-indexed trajectory; repeated runs can therefore reveal how often extreme queueing behavior occurs.
System dynamics aggregates populations into continuous stocks and flows, agent-based modeling simulates individuals making decisions, and discrete-event simulation moves entities through processes at scheduled state changes.
In discrete-event simulation, an event calendar makes execution efficient by skipping intervals in which the modeled state cannot change.
Activities are input service durations that are modeled independently of system congestion, while delays are state-dependent waits that emerge as simulation outputs.
Stochastic modeling often represents complexity rather than literal randomness: sampling from fitted distributions can approximate arrival and service variation without modeling every causal detail.
Simulation roles depend on the research question: passengers are entities in a customer-service model, while the employees serving them may be resources—or entities in a model of employee experience.

Chapters

0:00 Models Answer What-If Questions

A model is anything that helps answer a what-if question; a useful model makes predictions that align with reality.
Mental models draw on personal experience but can be difficult to explain or transfer to another person.
Quantitative models formalize assumptions so their predictions can be generated mechanically, by solving equations or running a simulation.

5:00 Solved Mathematical Models vs. Executed Simulations

Mathematical models can yield symbolic relationships; a projectile trajectory, for example, predicts when a ball returns to the ground by finding when its height reaches zero.
Simulation models are executed rather than solved, and each run produces one instance of system behavior.
Repeated simulation runs create data from which analysts infer patterns and make recommendations.

12:00 Static Outcomes and Dynamic Trajectories

A static model produces one outcome or a distribution of outcomes, as in a profit-risk model or a single answer sampled from a language model.
A dynamic model produces a time-indexed trajectory, such as airport security queue length throughout a day.
Repeated airport-model runs may produce different trajectories; an extreme 90-minute security wait matters partly because analysts need to estimate how often it occurs.

15:25 Building Dynamic Models from Inputs and State Updates

A dynamic model combines static input models, such as the distribution of time between airport arrivals, with rules that update system state.
State variables can include the number of people waiting for security and the number already in the gate area.
Arrival bursts generated by the input model can produce downstream queue buildup through the state-update rules.

18:45 Three Core Types of Dynamic Simulation

Dynamic models differ in what carries their state and how that state advances over time.
System dynamics, agent-based modeling, and discrete-event system simulation are presented as the three core approaches.
IEE 475 concentrates on discrete-event simulation; system dynamics is covered in IEE 477, and agent-based modeling appears in Lab 4.

23:20 System Dynamics: Aggregate Stocks and Continuous Flows

The Bass innovation-diffusion model tracks aggregate stocks such as potential adopters and adopters rather than individual customers.
Stock-and-flow diagrams represent state quantities and the differential-equation flows that transfer quantity between them over continuous time.
Epidemiological SIR models similarly track counts of susceptible, infectious, and recovered people, with infection flows depending on the current counts.

27:20 Agent-Based Models: Individual Decisions and Spatial Variation

Agent-based models represent individuals explicitly and commonly update them at fixed time steps, repeatedly asking agents whether they will act.
An airport example uses 496 passenger agents, random movement, and security lanes whose availability affects passenger progress.
Explicit spatial behavior can represent possibilities such as a forest fire spreading along different routes, but every modeled behavior must be justified and validated.

36:30 Discrete-Event Simulation: Jumping Between State Changes

Discrete-event simulation advances from one scheduled event to the next instead of computing every moment between events.
An event calendar or future-event list stores possible state-changing events, such as passenger arrivals and service completions.
Entities follow process logic rather than making autonomous decisions, enabling efficient simulation while retaining individual variation.

47:00 Arena Airport and Emergency-Room Examples

Arena models represent queues, resources, and work in progress, with visualizations that help stakeholders relate simulated behavior to familiar settings.
An emergency-room model routes patients according to condition severity and the availability of beds, doctors, and nurses.
Tracking resource use allows analysts to compare outcomes with costs, such as the expense of routing more patients to doctors.

51:00 Stochastic Models vs. Deterministic Computation

A stochastic model uses probability to simplify a complex system, even when the underlying process—such as a bank transaction—is not fundamentally random.
Analytic queueing methods solve formulas, while simulation samples from input distributions and analyzes the resulting runs.
Monte Carlo methods can also estimate deterministic quantities: Lab 3 uses random sampling to estimate π, although π itself is fixed.

57:00 System Boundaries, State Variables, and State

A system is a group of connected objects with a joint purpose, and its boundary determines which details—such as airport passengers’ appetite—must be modeled.
A state variable is a quantity such as classroom occupancy; its current value, such as 50 people, is the system’s state at that moment.
IEE 475 focuses on discrete-event systems whose state changes at events, rather than continuously changing quantities such as room temperature.

1:03:00 Resources, Entities, Activities, and Delays

Resources are limited-capacity model components, such as an airport X-ray scanner; entities such as passengers compete for access.
An activity is the time spent receiving service at a resource, while a delay is the state-dependent time spent waiting to access it.
The same object can be an entity or resource depending on the question: airport document checkers are resources in a passenger-service model but entities in an employee-experience model.

1:04:00 Activities Are Inputs; Delays Are System Outputs

Activity duration is modeled as an input that does not depend on how busy the system is; using an X-ray scanner takes its modeled service time either way.
Delay depends on system state: the time to reach a scanner is short when it is free and longer when a queue has formed.
Queue wait time should generally be measured as a simulation output, not supplied as an input distribution, because it emerges from arrivals and resource availability.

1:07:30 Entity Attributes, State Snapshots, and Flowchart Blocks

Manufacturing parts can carry attributes such as part number, color, type, and arrival time, allowing the model to calculate each part’s time in the system.
Machine busy/idle indicators can form a state vector; additional state variables may track queue lengths or resource availability.
Simulation flowcharts use creation blocks for arrivals, process blocks for activities and resource use, decision blocks for routing, and disposal blocks to record completed-entity statistics.

1:13:00 Randomized Arrival Inputs and the Closing Clicker Question

A deterministic arrival model might impose exactly 20 minutes between passengers, while a stochastic model samples intervals around a fitted distribution with a 20-minute average.
Random input models replace impractical rules based on every traveler’s circumstances; later lectures cover mapping random numbers to arrival times and outcomes such as an 80% security-pass rate.
The closing clicker checks the key terminology: time spent waiting to enter a resource is a delay, not an activity.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Ted Pavlic.