Index of /publis/slides/2019_04__Presentation_IEEE_WCNC__MoTION_Workshop

[ICO]NameLast modifiedSizeDescription
[PARENTDIR]Parent Directory  -  
[DIR]plots/2021-03-04 16:23 -  
[   ].gitignore2021-03-04 16:23 7  
[   ]Makefile2021-03-04 16:23 1.5K 
[TXT]README.md2021-03-04 16:23 10K 
[TXT]preprocess_tex.sh2021-03-04 16:23 119  
[TXT]slides.md2021-03-04 16:23 11K 
[   ]slides.pdf2021-03-04 16:23 1.6M 
[   ]slides.pdfpc2021-03-04 16:23 32  
[   ]slides_169.pdf2021-03-04 16:23 1.7M 
[   ]slides_169.pdfpc2021-03-04 16:23 36  
[   ]slides_pandoc.pdfpc2021-03-04 16:23 39  

title: Upper-Confidence Bound for Channel Selection in LPWA Networks with Retransmissions subtitle: MoTION Workshop @ IEEE WCNC 2019 author: Lilian Besson institute: SCEE Team, IETR, CentraleSupélec, Rennes date: Monday 14th of April, 2019

lang: english

1st MoTION Workshop - 2019: "Upper-Confidence Bound for Channel Selection in LPWA Networks with Retransmissions"

Christophe Moy
@ IETR, Rennes
Emilie Kaufmann
@ CNRS & Inria, Lille
8% 14% 12% 16%

See our paper at HAL.Inria.fr/hal-02049824


:timer_clock: Outline

1. Motivations

2. System model

3. Multi-armed bandit (MAB) model and algorithms

4. Proposed heuristics

5. Numerical simulations and results

Please :pray: ask questions at the end if you want!

By R. Bonnefoi, L. Besson, J. Manco-Vasquez and C. Moy.


1. Motivations

But...


2. System model

Wireless network

One gateway, many IoT devices


Transmission and retransmission model

The goal of each object

Is to maximize its successful communication rates $\Longleftrightarrow$ maximize its number of received Ack.


Do we need learning for transmission? Yes!

First hypothesis

The surrounding traffic is not uniformly occupying the $K$ channels.

Consequence


Do we need learning for retransmission?

Second hypothesis

Imagine a set of IoT devices learned to transmit efficiently (in the most free channel), in one IoT network.

Question


Mathematical intution and illustration

Consider one IoT device and one channel, we consider two probabilities:

In an example network with... - a small transmission probability $p=10^{-3}$, - from $N=50$ to $N=400$ IoT devices,


75%


Do we need learning for retransmission? Yes we do!

Consequence


3. Multi-Armed Bandits (MAB

3.1. Model

3.2. Algorithms


3.1. Multi-Armed Bandits Model

Why is it famous?

Simple but good model for exploration/exploitation dilemma.


3.2. Multi-Armed Bandits Algorithms

Often "index based"

Example: "Follow the Leader"


Upper Confidence Bounds algorithm (UCB)

Parameter $\alpha$: tradeoff exploration vs exploitation


Upper Confidence Bounds algorithm (UCB)

90%


4. We Study Different Heuristics (5)


4.0. Only UCB

Use the same $\mathrm{UCB}$ to decide the channel to use for any transmissions, regardless if it's a first transmission or a retransmission of a message.

80%


4.1. UCB + Random Retransmissions

90%


4.2. UCB + a single UCB for Retransmissions

90%


4.3. UCB + $K$ UCB for Retransmissions

85%


4.4. UCB + Random Retransmission

85%


5. Numerical simulations and results

What

Why


5.1. First experiment

We consider an example network with...

Non uniform occupancy of the $4$ channels: they are occupied $10$, $30$, $30$ and $30\%$ of times (by other IoT networks).


80%


5.2. Second experiment

Non uniform occupancy of the $4$ channels: they are occupied $40$, $30$, $20$ and $30\%$ of times (by other IoT networks).


80%


6. Summary (1/3)

Settings

  1. For LPWA networks based onan ALOHA protocol (slotted both in time and frequency),
  2. We presented a retransmission model
  3. Dynamic IoT devices can use simple machine learning algorithms, to improve their successful communication rate,
  4. We focus on the packet retransmissions upon radio collision, by using low-cost Multi-Armed Bandit algorithms, like UCB.

6. Summary (2/3)

We presented

Several learning heuristics


6. Summary (3/3)

We showed


6. Future works


More ?

→ See our paper: HAL.Inria.fr/hal-02049824

:pray: Please ask questions !

Thanks for listening !