samerlahoud / IoT-MAB

Decentralized Intelligent Resource Allocation for LoRaWAN Networks

IOT-MAB: Decentralized Intelligent Resource Allocation Approach for LoRaWAN Networks

Introduction

IoT-MAB is a discrete-event simulator based on SimPy for simulating intelligent distributed resource allocation in LoRa networks and to analyse scalability. We also combine the classed and functions for Physical layer of LoRA.

How to cite?

@misc{LoRa_MAB,
author =   {Duc-Tuyen Ta, Kinda Khawam, Samer Lahoud},
title =    {{LoRaWAN Network Simulator with Reinforcement Learning-based Algorithms}},
howpublished = {\url{https://github.com/tuyenta/IoT-MAB}},
}

Installation

It is recommend to use virtualenv to keep your Python environment isolated, together with virtualenvwrapper to make working with virtual environments much more pleasant, e.g.:

$ pip install virtualenvwrapper
...
$ export WORKON_HOME=~/.virtualenvs
$ mkdir -p $WORKON_HOME
$ source /usr/local/bin/virtualenvwrapper.sh
$ mkvirtualenv -p python3 iot_mab

You can install the required packages using the provided requirements.txt file:

(lorasim)$ pip install -r requirements.txt

Usage

Synopsis

python3 IoT_MAB.py <nrNodes> <nrIntNodes> <nrBS> <initial> <radius> <distribution> <AvgSendTime> <horizonTime>
<packetLength> <freqSet> <sfSet> <powerSet> <captureEffect> <interSFInterference> <infoMode> <logdir> <exp_name>

Example:

python3 IoT_MAB.py --nrNodes 5 --nrIntNodes 5 --nrBS 1 --initial UNIFORM --radius 2000 --distribution '0.1 0.1 0.3 0.4 0.05 0.05' --AvgSendTime 360000 --horizonTime 10  --packetLength 50 --freqSet '867300' --sfSet '7 8'  --powerSet "14"  --captureEffect 1  --interSFInterference 1 --infoMode NO --logdir logs --exp_name exp1

Description

nrNodes

number of nodes to simulate.

nrIntNodes

number of smart nodes to simulate. nrIntNodes must be smaller than nrNodes

nrBS

number of base station.

initial

initial probability for learning process, which is UNIFORM for uniform distribution or RANDOM for random distribution.

radius

radius to simulate in metre.

distribution

distribution of end-devices in the network

AvgSendTime

average sending interval in milliseconds.

horizonTime

number of iteration to simulate. The simulation time is horizonTime x AvgSendTime

packetLength

length of packet to simulate in bytes

sfSet

set of SF to simulate, must be between 7 and 12

freqSet

set of frequency to simulate.

powerSet

set of power to simulate.

captureEffect

capture effect (power collision) or not.

interSFInterference

inter-sf interference.

infoMode

information mode to simulate.

logdir

name of folder to store simulations.

exp_name

name of folder to store scenario.

Output

The result of every simulation run will be appended to a file named prob..._X.csv, ratio....csv, energy....csv and traffic....csv, whereby

• prob..._X is the probability of device X.

• ratio... is the packet reception ration of the network.

• energy... is the energy consumption of the network.

• traffic... is the normalized traffic and normalized throughput of the network.

The data file is then plotted into .png file by using matplotlib.

Changelogs

Contact

Duc-Tuyen Ta

Postdoc, ROCS, LRI, Paris-Sud University. [e-mail]

Kinda Khawam

Associate Professor at the University of Versailles. Associated to the ROCS team in LRI, Paris-Sud University. [e-mail]

Samer Lahoud

Faculté d’ingénierie ESIB, Université Saint-Joseph de Beyrouth, Lebanon [e-mail]