SIMULATION & MODELING (IT-702 (C)) - Unit 2 Short Notes

1.1 Introduction to Simulation

Definition & Purpose:

Simulation is the process of creating a computational model of a real-world system and conducting experiments with this model to understand system behavior or evaluate strategies. Its primary purpose is to analyze complex systems where analytical solutions are infeasible, enabling "what-if" analysis without disrupting the actual system.

Steps in a Simulation Study (Flow):

1. Problem Identification & Objectives: Define the issue and goals.

2. Conceptual Model Design: Outline system components, logic, and assumptions.

3. Data Collection & Input Modeling: Gather data, identify distributions.

4. Model Implementation: Code the model using a simulation language/tool.

5. Verification & Validation: Ensure model correctness and realism.

6. Experimentation & Output Analysis: Run simulations, analyze results statistically.

7. Documentation & Reporting: Present findings and recommendations.

[!TIP]

Exam Focus: The flow diagram is frequently asked. Sketch it as a cyclic process: Problem → Model Design → Data → Implementation → V&V → Experimentation → Results → (back to) Problem.

Simulation vs. Analytical Modeling:

Applications: Manufacturing (production lines), Healthcare (patient flow), Transportation (traffic networks), Logistics (supply chains), Defense (battlefield scenarios).

1.2 Types of Systems & Simulation

Discrete vs. Continuous Systems:

Stochastic vs. Deterministic:

• Stochastic: Incorporates randomness (e.g., random arrivals, service times). Uses random variables.

• Deterministic: No randomness; same input always yields same output (e.g., simple kinematic models).

Static vs. Dynamic:

• Static: Time is not a factor (e.g., Monte Carlo simulation for risk analysis).

• Dynamic: System evolves over time (most simulations are dynamic).

1.3 Probability & Statistics for Simulation

Random Variables (RVs):

• Discrete RV: Takes countable values (e.g., number of customers).

• Continuous RV: Takes any value in an interval (e.g., service time).

Key Probability Distributions:

1. Binomial Distribution (Discrete):

   Models number of successes in $n$ independent Bernoulli trials.

$$P(X = k) = \binom{n}{k} p^k (1-p)^{n-k}, \quad k=0,1,...,n$$

$$\displaystyle \mu = np $$, $$\displaystyle \sigma^2 = np(1-p) $$.

2. Poisson Distribution (Discrete):

   Models number of events in fixed interval when events occur independently at rate $\lambda$.

$$P(X = k) = \frac{\lambda^k e^{-\lambda}}{k!}, \quad k=0,1,2,...$$

$$\displaystyle \mu = \lambda $$, $$\displaystyle \sigma^2 = \lambda $$.

3. Normal Distribution (Continuous):

   Bell-shaped curve for natural phenomena.

$$f(x) = \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{(x-\mu)^2}{2\sigma^2}}, \quad -\infty < x < \infty$$

$\mu$ = mean, $\sigma$ = standard deviation.

[!TIP]

Exam Crucial: Binomial approximates Poisson when $n \to \infty$, $p \to 0$, and $$\displaystyle \lambda = np $$ remains constant (rare events). State this condition explicitly.

Density & Distribution Functions:

• PDF (Continuous): $f(x) \geq 0$, $$\displaystyle \int_{-\infty}^{\infty} f(x)dx = 1 $$. Probability $$\displaystyle P(a \leq X \leq b) = \int_a^b f(x)dx $$.

• PMF (Discrete): $$\displaystyle p(x) = P(X=x) $$, $$\displaystyle \sum_x p(x) = 1 $$.

• CDF: $$\displaystyle F(x) = P(X \leq x) $$. For continuous: $$\displaystyle F(x) = \int_{-\infty}^x f(t)dt $$.

Stochastic Variables: RVs representing random inputs in simulation (e.g., interarrival times, service times). Their distributions are identified from real data via input modeling.

1.4 Queuing Theory & Systems

General Structure (Kendall's Notation):
$A/B/c/K/N$

• $A$: Arrival distribution (M=Markovian/exponential, D=Deterministic, G=General)

• $B$: Service distribution

• $c$: Number of servers

• $K$: System capacity (max entities in queue + service)

• $N$: Population size (finite/infinite)

Example: $M/M/1/\infty/\infty$ → Poisson arrivals, exponential service, 1 server, infinite queue, infinite population.

Key Characteristics:

1. Arrival Pattern: Modeled by interarrival time distribution (often Poisson → exponential).

2. Service Mechanism: Number of servers, service time distribution.

3. Queue Discipline: FIFO (first-in-first-out), LIFO, priority, SIRO.

4. Capacity: Finite/infinite queue size.

Simulation of a Queuing System (Algorithm):

1. Initialize: clock=0, queue empty, server idle, statistics counters.

2. Schedule first arrival (generate from arrival distribution).

3. While (clock < simulation end time):

   a. Find next event (arrival or departure).

   b. Advance clock to event time.

   c. If arrival:

   • If server idle & queue empty → begin service, schedule departure.

   • Else → add to queue, update max queue length.

   d. If departure:

   • If queue non-empty → remove first from queue, start service, schedule departure.

   • Else → mark server idle.

   e. Update performance statistics (time in queue, server utilization, etc.).

4. After simulation, compute averages: $L$ (avg number in system), $$\displaystyle L_q $$ (avg queue length), $W$ (avg waiting time), $$\displaystyle W_q $$ (avg time in queue).

Performance Measures:

• Utilization ($\rho$): $$\displaystyle \rho = \lambda / (c\mu) $$ for $M/M/c$.

• $$\displaystyle L = L_q + \lambda / \mu $$ (Little's Law: $$\displaystyle L = \lambda W $$).

• $$\displaystyle W = W_q + 1/\mu $$.

1.5 Simulation Methodology

Verification vs. Validation:

Model Building:

• Conceptual Model: High-level diagram/logic of system components and interactions.

• Assumptions: Simplify reality (e.g., ignore breakdowns) while preserving essential behavior. Document all assumptions.

Input Data Analysis:

1. Identification: Determine which inputs are stochastic.

2. Collection: Gather historical data or use expert estimates.

3. Fitting: Use statistical tests (Chi-square, K-S) to fit distributions. Software: Stat::Fit, @RISK.

Output Analysis:

• Terminating Simulation: Runs for a fixed time or until a terminating event (e.g., single project).

• Steady-State Simulation: Runs long enough to reach equilibrium; requires warm-up period elimination.

• Use replication-deletion method for steady-state: discard initial warm-up data, average across replications.

• Statistical Inference: Compute confidence intervals for outputs (e.g., mean $$\displaystyle \bar{x} \pm t_{\alpha/2, n-1} \cdot s/\sqrt{n} $$).

1.6 Simulation Languages & Tools

Classification of Simulation Languages:

• General-Purpose: Can model any system (e.g., GPSS, Simscript).

• Special-Purpose: Tailored for specific domains (e.g., Arena for manufacturing, SimPy for Python-based DES).

• Simulation of Classification Languages: Languages designed to model classification problems (e.g., decision trees) within simulation environments for AI-driven decision logic.

Popular Simulation Software:

• Arena: GUI-based DES, integrates with Excel, strong in manufacturing/logistics.

• SimPy: Open-source, Python-based, process-oriented DES.

• AnyLogic: Multi-method (DES, ABS, SD), Java-based, supports 3D visualization.

• MATLAB/Simulink: Continuous and discrete dynamic systems, control systems.

1.7 Advanced Simulation Concepts

Analog vs. Digital Simulation:

• Analog: Uses continuous physical phenomena to model system (e.g., electrical circuits for mechanical systems). Rare today.

• Digital: Uses discrete computational steps (all modern software-based simulation).

AI Techniques in Simulation:

• Neural Networks: Approximate complex input-output relationships, optimize parameters.

• Fuzzy Logic: Handle vague inputs (e.g., "high traffic"), useful in control systems.

• Genetic Algorithms: Optimize simulation parameters (e.g., find best routing policy).

Pure Pursuit Problem:

A path-tracking algorithm where a vehicle follows a reference path by steering toward a lookahead point at a fixed distance ahead. Used in autonomous vehicle simulation. Key parameters: lookahead distance, vehicle kinematics.

Autopilot Simulation:

Example of complex dynamic system simulation. Components:

1. Sensors: GPS, IMU, air data.

2. Guidance Law: Computes desired trajectory (e.g., pure pursuit).

3. Control System: PID/adaptive control to track trajectory.

4. Actuators: Ailerons, rudder, throttle.

Simulated in tools like MATLAB/Simulink with 6-DOF aircraft models.

1.8 Limitations & Challenges of Simulation

1. Computational Cost: Large-scale models require significant time/resources (especially steady-state with many replications).

2. Model Complexity: Overly detailed models become unmanageable; balance realism with simplicity.

3. Validity & Credibility: Model may not accurately represent reality; requires thorough V&V.

4. Random Number Quality: Poor RNGs lead to biased results; use tested RNGs (e.g., Mersenne Twister).

5. Output Analysis Difficulty: Distinguishing signal from noise requires statistical expertise.

6. Cost of Model Development: Skilled personnel and time-intensive.

[!TIP]

Common Pitfall: Students often state "simulation is expensive" without specifying why. Always link to computational cost, development time, or V&V effort.

Cross-Cutting Synthesis (For UNIT 2):

• Verification & Validation is the cornerstone of credible simulation, applicable to all models (queuing, VR, Big Data-driven).

• Stochastic inputs (from probability distributions) drive DES; their quality impacts output reliability.

• Performance measures (e.g., $L$, $W$ in queues) are analogous to QoS metrics in cloud or frame rates in VR—all quantify system effectiveness.

• Simulation workflow (Problem → Model → Data → Experiment) is universal; only tools and complexity vary by domain (Paper A/B/D).