Skip to content

simulation_presentation.md

Latest commit

 

History

History
159 lines (112 loc) · 3.8 KB

simulation_presentation.md

File metadata and controls

159 lines (112 loc) · 3.8 KB
title Simulation Framework & Scenarios
subtitle Discrete Event Simulator Documentation
author Sergio Gimenez
date December 4, 2025
theme CambridgeUS
colortheme default
header-includes
\usepackage{ulem}
\usepackage{amsmath}
\usepackage{fontspec}
\usepackage{unicode-math}
\newfontfamily\DejaSans{DejaVu Sans}
\let\emojifont\DejaSans

Agenda

  1. How the Simulator Works (In a Nutshell)
  2. Available Scenarios
  3. Single-Tier Scenario Setup
  4. Two-Tier Scenario Setup

1. How the Simulator Works

The goal is to generate a synthetic but realistic network graph.

Steps:

  1. Load gNBs: Real data from OpenCellID.
  2. Load Agents: Synthetic users based on population density.
  3. Distribute UPFs: Optimal placement using K-Means.
  4. Connect the Dots: Proximity-based connections.

Step 1: Loading gNBs

  • Source: OpenCellID (Public dataset).
  • Data: Latitude and Longitude of cell towers.
  • Goal: Get ballpark numbers and realistic distribution of Base Stations.

\begin{center} \includegraphics[width=0.45\textwidth]{docs/simulation-details/in-a-nutshell/images/topology_spain_movistar_gnbs.png} \end{center}

Step 2: Loading Agents

  • Agents: Represent groups of users (e.g., 1 agent = 1000 users).
  • Distribution: Proportional to municipality population.
  • Constraints: Placed within valid geographic boundaries (polygons) of municipalities.

\begin{center} \includegraphics[width=0.45\textwidth]{docs/simulation-details/in-a-nutshell/images/topology_spain_movistar_agents.png} \end{center}

Step 3: Distributing UPFs

  • Goal: Place UPFs to minimize squared distance to gNBs.
  • Method: K-Means Clustering.
  • Optimization: Purely geometric (ignoring orography/roads for simplicity).
  • Configurable: The number of UPFs is a parameter (e.g., 52 for Spain - one per province).

Step 4: Connecting the Dots

We build the graph based on proximity (minimum Haversine distance).

graph LR
    UE([UE]) -- NR (closest) --> gNB[gNB]
    gNB -- N3 (closest) --> UPF[UPF]
Loading

\begin{center} \includegraphics[width=0.6\textwidth]{docs/scenarios/single_tier_architecture/images/graph_viz_movistar_spain distributed.png} \end{center}

2. Available Scenarios

We have identified three main simulation scenarios:

  1. Single-Tier Architecture: Baseline flat topology.
  2. Two-Tier Architecture: Hierarchical topology with Edge and Regional UPFs.
  3. Mobility: Scenarios involving user movement and handovers.

3. Single-Tier Scenario Setup

Topology:

  • Flat network structure.
  • gNBs connect directly to distributed UPFs.
  • UPFs are independent (no interconnection).
graph LR
    UE([UE]) -- NR --> gNB[gNB]
    gNB -- N3 --> UPF[UPF]
Loading

Single-Tier Configuration

Config File (config.toml):

[simulation]
scenario_mode = "single_tier"

Scenario Definition:

  • Define the number of UPFs to spread across the country.
  • Example: "Spain Edge" = 52 (One UPF per province).

4. Two-Tier Scenario Setup

Topology:

  • Hierarchical structure.
  • Edge UPFs (UL-CL): Close to gNBs, aggregate traffic.
  • Centralized UPFs (PSA): Regional anchors.
graph LR
    UE([UE]) -- NR --> gNB([gNB])
    gNB -- N3 --> EdgeUPF(["Edge UPF"])
    EdgeUPF -- N9 --> PSA(["PSA UPF"])
Loading

Two-Tier Key Differences

Hierarchical K-Means Clustering:

  1. Level 1: Cluster gNBs to place Edge UPFs (e.g., 52 locations).
  2. Level 2: Cluster Edge UPF locations to place Centralized UPFs (e.g., 5 locations).

Configuration:

[simulation]
scenario_mode = "two_tier"
num_centralized_upfs = 5
  • The scenario number (e.g., 52) now refers to the count of Edge UPFs.