<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<p>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
</p>
<div class="text_exposed_show">
<p><br>
</p>
<p align="center">Tutorial <br>
</p>
<p align="center">The Evolutionary Robotics Toolbox for
Developmental Robotics<br>
</p>
<p align="center">Monday 19th September, afternoon session
(14:00-17:00)</p>
<p align="center"> IEEE-ICDL Cergy, France
(<a class="moz-txt-link-freetext" href="http://www.icdl-epirob.org/">http://www.icdl-epirob.org/</a>)<br>
</p>
<p><br>
</p>
<p> More details here:
<a class="moz-txt-link-freetext" href="http://pages.isir.upmc.fr/~bredeche/evorobots_tutorial/">http://pages.isir.upmc.fr/~bredeche/evorobots_tutorial/</a></p>
<p><br>
</p>
<p>*Abstract:*<br>
Evolutionary robotics (ER) started in the nineties with the
objective of using evolution-inspired algorithms to design
robots: since natural evolution shaped all the lifeforms we
know, could artificial evolution shape better robots? After 25
years, this general question led to many general insights about
search and adaptation. For instance, it is now well established
in ER that searching for novel behaviors can be more effective
than optimizing an objective (Lehman and Stanley, 2011), that
encouraging exploration in the behavior space is more effective
than encouraging it in the genotypic space (Mouret and Doncieux,
2012), or that behaviors discovered in simulation are unlikely
to work on real robots (Jakobi, 1997).</p>
<p> In spite of the different starting points of Developmental
Robotics and ER, many recent ideas introduced in the two fields
are surprisingly similar. For example, open-ended evolution
(e.g. Novelty Search) parallels intrinsic motivation in
developmental robotics (Oudeyer and Kaplan, 2007), the behavior
space in ER is similar to the goal space in Developmental
Robotics (Rolf et al. 2010), and the relationships between
simulation and reality is linked to self-modeling (Bongard et
al., 2006). </p>
<p> Inspired by these parallels, the main goal of this tutorial is
to introduce recent ideas from ER that can be interesting for
Developmental Robotics. Over the course of this half-day, the
instructors will discuss the divergences and the convergences
between the two fields, as well as the common challenges.</p>
<p> *Organizers*<br>
Stéphane Doncieux, ISIR, UPMC, France.
<a class="moz-txt-link-abbreviated" href="mailto:stephane.doncieux@isir.upmc.fr">stephane.doncieux@isir.upmc.fr</a> <br>
Nicolas Bredeche, ISIR, UPMC, France,
<a class="moz-txt-link-abbreviated" href="mailto:nicolas.bredeche@isir.upmc.fr">nicolas.bredeche@isir.upmc.fr</a> <br>
Jean-Baptiste Mouret, INRIA Nancy, France,
<a class="moz-txt-link-abbreviated" href="mailto:jean-baptiste.mouret@inria.fr">jean-baptiste.mouret@inria.fr</a></p>
<p> *Table of contents*<br>
- Introduction<br>
- Evolutionary algorithms (EA) for learning & babbling
(single individual)<br>
- Interactions between simulation and reality<br>
- pause<br>
- Selective pressures<br>
- Convergent vs divergent search<br>
- Evolutionary Algorithms for group dynamics<br>
- Conclusion & questions</p>
</div>
<br>
<br>
<br>
<pre class="moz-signature" cols="72">--
Stephane Doncieux
Professeur/Professor
ISIR - UPMC/CNRS
Pyramide Tour 55 Boite courrier 173
4 place Jussieu 75252 Paris cedex 05
France
Tel: +33 1 44 27 87 45
Web: <a class="moz-txt-link-freetext" href="http://people.isir.upmc.fr/doncieux">http://people.isir.upmc.fr/doncieux</a>
</pre>
</body>
</html>