Blender: Toward Practical Simulation Framework for BLE Neighbor Discovery

Overview

Blender is an open-source simulation framework for Bluetooth Low Energy (BLE) neighbor discovery, built on top of the NS-3 network simulator. It provides accurate modeling of the BLE 5.0 protocol stack, enabling researchers to evaluate and prototype neighbor discovery schemes without costly hardware deployments.

Motivation

The Challenge

BLE neighbor discovery is a fundamental building block for many IoT applications, including:

Contact tracing
Asset tracking
Proximity-based services
Device pairing

However, evaluating neighbor discovery schemes presents significant challenges:

High Cost: Real-world deployments require multiple devices and extensive testing
Limited Reproducibility: Environmental factors make experiments difficult to reproduce
Time-Consuming: Large-scale studies can take weeks or months
Lack of Accuracy: Existing simators oversimplify BLE protocol behavior

Our Solution

Blender addresses these challenges by providing a practical, accurate, and efficient simulation framework validated against real-world experiments.

System Design

Architecture

Blender extends NS-3 with a comprehensive BLE 5.0 implementation:

┌─────────────────────────────────────┐
│     Application Layer               │
│   (Neighbor Discovery Logic)        │
├─────────────────────────────────────┤
│     BLE Protocol Stack              │
│  - Link Layer                       │
│  - PHY Layer (BLE 5.0)             │
├─────────────────────────────────────┤
│     NS-3 Core                       │
│  - Mobility Models                  │
│  - Channel Models                   │
│  - Interference                     │
└─────────────────────────────────────┘

Key Features

1. Accurate BLE 5.0 Modeling

Implemented Features:

Complete BLE 5.0 link layer state machine
Advertising states (ADV_IND, ADV_DIRECT_IND, etc.)
Scanning states (Passive, Active)
Connection establishment procedures
Channel selection algorithm (37 channels: 3 advertising + 34 data)

Real-World Factors:

Duty cycling
Clock drift
Radio propagation delays
Interference from other devices

2. Configurable Parameters

Parameter	Range	Default
Advertising Interval	20ms - 10.24s	100ms
Scan Interval	10ms - 10.24s	100ms
Scan Window	10ms - 10.24s	50ms
TX Power	-20dBm - +10dBm	0dBm
RX Sensitivity	-100dBm - -50dBm	-80dBm

3. Multiple Discovery Schemes

Supported Schemes:

BLE Standard: Legacy BLE discovery
BleDiCo: Directed discovery with beacons
DiffBlue: Differential-based discovery
Custom: User-defined schemes

Validation

Experimental Setup

We validated Blender against real-world deployments:

Testbed:

10 TI CC2650 devices
3 different environments (office, hallway, outdoor)
100+ discovery events per configuration

Metrics:

Discovery latency
Energy consumption
Success rate

Results

Accuracy

Metric	Real World	Blender	Error
Discovery Time (avg)	2.3s	2.28s	< 1%
Energy per Discovery	45mJ	44mJ	2.2%
Success Rate	94%	93%	1.1%

Overall validation error: < 5% ✓

Scalability

Blender can simulate large-scale scenarios efficiently:

Devices	Simulation Time	Memory Usage
10	2 min	50 MB
100	15 min	200 MB
1000	2 hours	1.5 GB

Case Studies

Case Study 1: Impact of Density

Setup: Vary number of devices from 10 to 1000

Findings:

Discovery time increases logarithmically with density
Energy consumption increases linearly due to more collisions
Optimal advertising interval: 200-300ms for dense scenarios

Case Study 2: Heterogeneous Intervals

Setup: Mix of devices with different advertising/scan intervals

Findings:

Heterogeneity significantly impacts discovery latency
Worst-case: 5x increase when mixing fast and slow devices
Recommendation: Synchronize intervals or use adaptive schemes

Case Study 3: Mobile Scenarios

Setup: Random waypoint and real human mobility traces

Findings:

Mobility increases discovery opportunities
Contact duration becomes critical
Need for rapid discovery in mobile scenarios

Usage

Installation

# Clone Blender repository
git clone https://github.com/blender-ble/blender-ns3.git
cd blender-ns3

# Build NS-3 with Blender
./ns3 build

Example Script

import ns.core
import ns BLE
import ns.internet
import ns.mobility

# Create nodes
nodes = ns.network.NodeContainer()
nodes.Create(10)

# Install BLE devices
bleHelper = ns.BLE.BleHelper()
bleDevices = bleHelper.Install(nodes)

# Configure advertising
bleHelper.SetAdvertisingInterval(100)  # ms
bleHelper.SetScanInterval(100)  # ms
bleHelper.SetScanWindow(50)  # ms

# Setup mobility
mobility = ns.mobility.MobilityHelper()
mobility.SetMobilityModel("ns3::RandomWaypointMobilityModel",
                          "Speed", ns.core.StringValue("ns3::UniformRandomVariable[Min=1.0|Max=2.0]"),
                          "Pause", ns.core.StringValue("ns3::ConstantRandomVariable[Constant=0.0]"))
mobility.Install(nodes)

# Run simulation
ns.core.Simulator.Stop(ns.core.Seconds(100))
ns.core.Simulator.Run()

Impact

Adoption:

Used by 10+ research groups worldwide
Cited in 20+ papers on BLE and IoT
Integrated into teaching curricula

Benefits:

Cost Reduction: Eliminates need for expensive testbeds
Faster Iteration: Enables rapid prototyping
Better Science: Improves reproducibility of research

Publication

Blender: Toward Practical Simulation Framework for BLE Neighbor Discovery

Yuxia Ding, Tong Li, Jiaxin Liang, Dong Wang

Proceedings of the 25th International ACM Conference on Modeling Analysis and Simulation of Wireless and Mobile Systems (MSWiM ‘22), October 2022

DOI: 10.1145/3551659.3559052

PDF

Code

Future Work

BLE 5.2+ Support: Add support for latest BLE features
Energy Modeling: More detailed energy consumption models
Machine Learning: Integration with ML-based optimization
Web Interface: Online simulation platform