In this tutorial, the use of evolutionary algorithms in robotics is studied. In view of their
wide-ranging applicability and their ability to search efficiently through large and complex search spaces, evolutionary algorithms are becoming increasingly important as optimization tools in many fields of science and technology, including robotics. The tutorial begins with two illustrative and pedagogical examples, one involving a simulated robot and the other a small experimental robot. A thorough introduction to evolutionary algorithms is then given, starting with basic concepts and proceeding to complex topics, such as the use of evolutionary methods in open-ended problems, where the size of the system being optimized is not known in advance. Next, in the main part of the tutorial, evolutionary robotics is studied from a variety of aspects through a set of illustrative examples, many involving robotics simulation software. Both simple examples, involving a single behavior, and complex examples, involving a repertoire of behaviors (and thus also involving the important problem of how to organize behaviors) are considered. Commercial applications (both present and future) will be studied as well.

Copies of all the software used during the tutorial will be given to the participants. In addition, the participants will be given a complete set of tutorial notes (on a CD).

Dr. Mattias Wahde is associate professor at the Department of Applied Mechanics at Chalmers University of Technology in Goteborg, Sweden. He is currently advising several PhD students and around 20 master students, and is also involved in collaborative research efforts with groups in Japan and Denmark. He also teaches several courses including a course in adaptive algorithms and intelligent machines, for which he has been awarded two pedagogical prizes.
Complete CV, see http://www.me.chalmers.se/~mwahde/CV.pdf

Dr. Jimmy Pettersson received his PhD degree from Chalmers University of Technology in 2006, with a thesis entitled “Generation and Organization of Behaviors for Autonomous Robots”. He is currently doing a postdoc in Japan, funded by the Canon Foundation.
Complete CV, see http://www.mvs.chalmers.se/~jimmyp/docs/jimmy_cv.pdf

The research interests of both speakers are focused on biologically inspired computation methods (such as e.g. evolutionary algorithms) as well as their applications in several fields, mainly robotics. Within this field, the organizers’ main area of interest is the problem of behavioral organization in autonomous robots, but they also work on robotic navigation, gait evolution for bipedal walking robots, fundamental properties of evolutionary algorithms etc.

The main motivation for the tutorial is the fact that evolutionary algorithms (e.g. genetic algorithms), being very powerful methods for search and optimization, are becoming increasingly widespread in many fields of science, including robotics. Thus, knowledge of such algorithms is essential. However, optimal use of such algorithms in connection with robotics requires significant experience, and is thus far from trivial. The main objectives with the tutorial is to give the participants
(1) Knowledge and understanding of the functionality, advantages, and disadvantages of evolutionary algorithms in connection with various robotics applications.
(2) Detailed knowledge of some of the more advanced techniques that are especially useful in robotics applications, such as e.g. using dynamically sized chromosomes for unrestrained optimization, using evolutionary algorithms for simultaneous optimization of the structure and function of a robot, and using evolutionary methods to solve the problem of behavioral organization in behavior-based robotics.
(3) Several illustrative examples of applications of evolutionary methods in robotics. The examples will be presented using robotics simulation software that will be given to the participants of the tutorial. The examples will include both research applications and industrial applications (present and future).
(4) A large and annotated set of relevant references (in the tutorial notes) to the topics discussed during the tutorial.

1. Introduction to evolutionary robotics
  1.1 Biological and artificial evolution
  1.2 Autonomous robots
  1.3 Examples of evolved behaviors
    1.3.1 Example 1: Simulation example: cleaning behavior
    1.3.2 Example 2: Applied example: cleaning behavior in Khepera.
2. Fundamentals of evolutionary methods
  2.1 Introduction
  2.2 Main components of evolutionary algorithms
  2.3 Basic applications of evolutionary algorithms: some examples
  2.4 Different versions of evolutionary algorithms
  2.5 Advanced topics
    2.5.1 Open-ended evolution
    2.5.2 Optimal parameter selection in evolutionary algorithms
3. Evolutionary robotics
  3.1 Introduction
  3.2 Evolving basic behaviors
  3.3 Behavioral organization and selection
  3.4 Case study 1: Robust navigation in changing environments
  3.5 Case study 2: Cleaning and garbage collection
  3.6 Case study 3: Balancing and walking
  3.7 Case study 4: A complete transportation robot
  3.8 The future of evolutionary robotics


Notes:

(1) The tutorial is being modified at present, and some modifications may therefore be made. Also, the number of case studies in part 3 may increase.
(2) Case study 4 (“A complete transportation robot”) will be particularly interesting for industry representatives, as it will describe, in detail, the development, using evolutionary methods, of a behavior-based architecture for realistic applications (in this case a transportation robot, but the description will be sufficiently general to be applicable to other cases. Basically, the example will show how the difficult problem of behavioral organization, i.e. electing (at all times) the most appropriate behavior, can be solved.
(3) The examples and case studies will be thoroughly illustrated with pictures, moving images, simulation software, and (in some cases) actual robots.