1 of 28

A mobile robotic platform for rapid prototyping of dynamic neural network models

(Master Defense)

Santino NANINI

Mazel

2 of 28

Summary

  1. Complex systems behaviour and robots dynamics

  • How to control/build a robot from scratch?

3 of 28

First part:

Complex systems behaviour

and robots dynamics

4 of 28

Metronomes synchronization experiment

5 of 28

Oscillators synchronization

Disorganized

time

coupled/ in same phase

6 of 28

From metronomes to neuron networks

7 of 28

General biological systems synchronization

Different types of “oscillators”

8 of 28

θi= 1

θj= 0

t= 0

9 of 28

θi= 1

θj= 0

t= 0

Coupling/Synchronization process

10 of 28

Random

θi= 1

θj= 0

θk= a

t= 0

t= T

Coupling/Synchronization process

All oscillators coupled

11 of 28

NOW IT’S YOUR TURN:

Please go to the link I sent you

12 of 28

First part “conclusion”:

Robots help to understand how the system is synchronizing, until the oscillators are all coupled together

Coupling/synchronization process analysis

13 of 28

Part two

How to build and control a robot from scratch?

14 of 28

  • Avoid very complex simulation of an environment
    • Real time environment acquisition via smartphone captors “grafted” to the robot

  • We want synthetic neural networks to learn how to develop synthetic perception + cognition

  • Robot = embodiment of synthetic neural networks

Motivation: why a robot?

15 of 28

Learning from “scratch”

I am a biologist,

I did not know much about

computer sciences

and nothing about electronics ...

16 of 28

ARDUINO

17 of 28

Arduino is cool . Proof:

18 of 28

Digital pin:

Send or receive HIGH (1) +5V

or LOW (0) 0V signal

Ex: LED, contact (press-on/ press- off button),

“All or nothing signal

Analog pin:

Send or receive a multitude of values as signal

And convert this values in a set of values between 0 and 1023

Ex: temperature, light-intensity...

19 of 28

DC Motors Speed control

20 of 28

H Bridge - “switcher”

Go!

Stop!

Electromagnetic brakes

Sens 1

Sens -1

21 of 28

Double H bridge - L293N

General connections

L293N organization

22 of 28

Photoresistance

  • Luminosity high resistance low voltage high
  • Luminosity low resistance high voltage low

23 of 28

Mechanism overview

TURN

TO THE LEFT

24 of 28

Things I had the chance to discover :

-Python in the beginning (as “starter”)

- electronics (Arduino)

- C/C++ mix (for Arduino)

- HTML, CSS,

Javascript (in progress)

25 of 28

Thanks to ...

Roberto TORO

Katja HEUER

Liubov TUPIKINA

Kevin LHOSTE

Daniel ASSAYAG

CRI!!!

26 of 28

Thank you

Danke

Shukria

Shukran

Merci

Toda

Questions?

27 of 28

Kuramoto model (1975)

Natural frequency

Number oscillators

Phase changemement other time

Coupling constant

Describes phase (angle) change other time, in N coupled oscillators

UNIVERSAL MODEL

Problem:

if N is high, then synchronization is very complex

-> we need to run simulation

28 of 28

i/dt = ωi + K/3 . (sin (θpi) + sin (θji))

p/dt = ωi + K/2 . (sin (θip))

j/dt = ωi + K/2 . (sin (θji))

Coupled to p and j

to i

to i

i

p

j

Kuramoto example