EvoROB2003
2nd European Workshop on Evolutionary Robotics

Previous editions:
Paris, France, 1998

Introduction

EvoROB2003 is the second workshop of the EvoNet working group on evolutionary robotics.

Evolutionary Robotics (ER) is a field of research that is concerned with the use of evolutionary computing techniques for the automatic design of adaptive robots. The aims of this workshop, which will bring together active ER researchers and people from industry, are to assess the current state-of-the-art and to provide opportunities for fostering future developments and applications. Non-European contributors or attendees are welcome.

Topics of interest include, but are not limited to:

• Genotype to Phenotype mappings
• Analysis of evolutionary processes and of evolved robots
• Adaptation to changing environments, to breakdowns or to material wear
• Co-evolution of control architectures and body plans
• Interactions of evolution, development and learning
• Evolvable Hardware in Robotics
• Fundamental methodological issues
• Comparison of Evolutionary Robotics methodologies
• Simulation-reality transferrence
• Scaling to complex behaviors
• Role of Evolutionary Robotics in Cognitive Science
• Neuroethological uses of Evolutionary Robotics

Papers are expected to deal with research involving real robots or with simulation experiments that explicitly deal with issues of relevance to real robots. The workshop proceedings will be published by Springer in the LNCS series and will be available at the workshop.

Draft: subject to change

See also: Programme overview

The EvoWorkshops2003 proceedings will be
published by Spinger as part of their
Lecture Notes in Computer Science series.

Behavioral plasticity in autonomous agents: a comparison between two types of controller
Tuci E, Quinn M
Blynel et al. recently compared two types of recurrent neural network, Continuous Time Recurrent Neural Networks (CTRNNs) and Plastic Neural Networks (PNNs), on their ability to control the behaviour of a robot in a simple learning task; they found little difference between the two. However, this may have been due to the simplicity of their task. Our comparison on a slightly more complex task yielded very different results: 70% runs with CTRNNs produced successful learning networks; runs with PNNs failed to produce a single success.
EvoROB Session 3: : April 15, 1345-1515

Competitive Co-Evolution of Predator and Prey Sensory-Motor Systems
Buason G, Ziemke T
A recent trend in evolutionary robotics research is to maximize selforganization in the design of robotic systems in order to reduce the human designer bias. This article presents simulation experiments that extend Nolfi and Floreano’s work on competitive co-evolution of neural robot controllers in a predator-prey scenario and integrate it with ideas from work on the ‘coevolution’ of robot morphology and control systems. The aim of the twenty-one experiments summarized here has been to systematically investigate the tradeoffs and interdependencies between morphological parameters and behavioral strategies through a series of predator-prey experiments in which increasingly many aspects are subject to self-organization through competitive co-evolution. The results illustrate that competitive co-evolution has great potential as a method for the automatic design of robotic systems.
EvoROB Session 2: : April 15, 1120-1250

Evolution of Collective Behavior in a Team of Physically Linked Robots
Baldassarre G, Nolfi S, Parisi D
In this paper we address the problem of how a group of four assembled simulated robots forming a linear structure can co-ordinate and move as straight and as fast as possible. This problem is solved in a rather simple and effective way by providing the robots with a sensor that detects the direction and intensity of the traction that the turret exerts on the chassis of each robot and by evolving their neural controllers. We also show how the evolved robots are able to generalize their ability in rather different circumstance by: (a) producing coordinated movements in teams with varying size, topology, and type of links; (b) displaying individual or collective obstacle avoidance behaviors when placed in an environment with obstacles; (c) displaying object pushing/pulling behavior when connected to or around a given object.
EvoROB Session 1: : April 15, 1000-1100

Evolving Motion of Robots with Muscles
Mahdavi S, Bentley P
The objective of this work is to investigate how effective smart materials are for generating the motion of a robot. Because of the unique method of locomotion, an evolutionary algorithm is used to evolve the best combination of smart wire activations to move most efficiently. For this purpose, a robot snake was built that uses Nitinol wire as muscles in order to move. The most successful method of locomotion that was evolved, closely resembled the undulating motion of the cobra snake. During experimentation, one of the four Nitinol wires snapped, and the algorithm then enabled adaptive behaviour by the robot by evolving another sequence of muscle activations that more closely resembled the undulations exhibited by the earthworm.
EvoROB Session 3: : April 15, 1345-1515

Evolving Neural Networks for the Control of a Lenticular Blimp
Doncieux S, Meyer J
We used evolution to shape a neural controller for keeping a blimp at a given altitude, and as horizontal as possible, despite disturbing winds. The blimp has a lenticular shape whose aerodynamic properties make it quite different from a classical cigar-shaped airship. Evolution has exploited these features to generate a neural network that proved to be more efficient than a hand-designed PID-based controller that independently controlled the blimp's three degrees of freedom.
EvoROB Session 3: : April 15, 1345-1515

Evolving Spiking Neuron Controllers for Phototaxis and Phonotaxis
Damper R, French R
Our long-term goal is to evolve neural controllers which reproduce in behaving robots the kind of phonotaxis behaviour seen in real animals, such as crickets. We have previously studied the evolution of neural circuitry which, when implanted in a Braitenberg type~2b vehicle, promoted phototaxis behaviour in the form of movement towards flashing lights of different frequencies. (It was simpler to study light-driven than acoustic-driven behaviour.)\ \ Since this is not truly sequential behaviour, we now describe new work to discriminate between particular mark-space ratio patterns of the same basic (flash or on-off) frequency. The next step will be to integrate the two behaviours so that robot taxis is driven by a signal with temporal structure closer to that of the cricket `song'.
EvoROB Session 2: : April 15, 1120-1250

Evolving Symbolic Controllers
Godzik N, Schoenauer M, Sebag M
The idea of symbolic controllers tries to bridge the gap between the top-down manual design of the controller architecture, as advocated in Brooks' subsumption architecture, and the bottom-up designer-free approach that is now standard within the Evolutionary Robotics community. The designer provides a set of elementary behavior, and evolution is given the goal of assembling them to solve complex tasks. Two experiments are presented, demonstrating the efficiency and showing the recursiveness of this approach. In particular, the sensitivity with respect to the proposed elementary behaviors, and the robustness w.r.t. generalization of the resulting controllers are studied in detail.
EvoROB Session 2: : April 15, 1120-1250

Exploring the T-Maze: Evolving Learning-Like Robot Behaviors using CTRNNs
Blynel J, Floreano D
This paper explores the capabilities of continuous time recurrent neural networks (CTRNNs) to display reinforcement learning-like abilities on a set of T-Maze and double T-Maze navigation tasks, where the robot has to locate and “remember” the position of a reward-zone. The “learning” comes about without modifications of synapse strengths, but simply from internal network dynamics, as proposed by [12]. Neural controllers are evolved in simulation and in the simple case evaluated on a real robot. The evolved controllers are analyzed and the results obtained are discussed.
EvoROB Session 1: : April 15, 1000-1100

Chairs

Agnès Guillot, Université Paris 6, France <agnes.guillot@lip6.fr >
Jean-Arcady Meyer, Université Paris 6, France <jean-arcady.meyer@lip6.fr>

Programme committee

• Wolfgang Banhaf, University of Dortmund, Germany
• Marco Dorigo, Université Libre de Bruxelles, Belgium
• Dario Floreano, EPFL, Switzerland
• Takashi Gomi, AAI, Canada
• John Hallam, University of Edinburgh, UK
• Inman Harvey, University of Sussex, UK
• Patrick Hénaff, Université de Versailles, France
• Phil Husbands, University of Sussex, UK
• Auke Jan Ijspeert, EPFL, Switzerland
• Pier Luca Lanzi, Politecnico di Milano, Italy
• Enrik Hautop Lund, University of Aarhus, Denmark
• Stefano Nolfi, National Research Council, Italy
• Peter Nordin, Chalmers University, Sweden
• Rolf Pfeifer, University of Zürich, Switzerland
• Olivier Sigaud, Université Paris 6, France
• Tom Ziemke, University of Skövde, Sweden