This paper examines the research paradigm of neurally-driven Evolutionary Autonomous Agents (EAAs), from a neuroscience perspective. Two fundamental questions are addressed: 1. Can EAA studies shed new light on the structure and function of biological nervous systems? 2. Can these studies lead to the development of new neuroscienti c analysis tools? The value and signi cant potential of EAA modeling in both respects is demonstrated and discussed. While the study of EAAs as a neuroscience research methodology still faces dicult conceptual and technical challenges, it is a promising and timely endeavor.

Abstract. It is clear to all, after a moments thought, that nature has much we might be inspired by when designing our systems, for example: robustness, adaptability and complexity, to name a few. The implementation of bio-inspired systems in hardware has however been limited, and more often than not been more a matter of artistry than engineering. The reasons for this are many, but one of the main problems has always been the lack of a universal platform, and of a proper methodology for the implementation of such systems. The ideas presented in this paper are early results of a new research project, "Reconfigurable POEtic Tissue". The goal of the project is the development of a hardware platform capable of implementing systems inspired by all the three major axes (phylogenesis, ontogenesis, and epigenesis) of bio-inspiration, in digital hardware. 1

This paper is concerned with adaptation capabilities of evolved neural controllers. We propose to evolve mechanisms for parameter self-organization instead of evolving the parameters themselves. The method consists of encoding a set of local adaptation rules that synapses follow while the robot freely moves in the environment. In the experiments presented here, the performance of the robot is measured in environments that are different in significant ways from those used during evolution. The results show that evolutionary adaptive controllers solve the task much faster and better than evolutionary standard fixed-weight controllers, that the method scales up well to large architectures, and that evolutionary adaptive controllers can adapt to environmental changes that involve new sensory characteristics (including transfer from simulation to reality and across different robotic platforms) and new spatial relationships.

For many years, researchers in the field of mobile robotics have been investigating the use of genetic and evolutionary computation (GEC) to aid the development of mobile robot controllers. Alongside the fundamental choices of the GEC mechanism and its operators, which apply to both simulated and physical evolutionary robotics, other issues have emerged which are specific to the application of GEC to physical mobile robotics. This paper presents a survey of recent methods in GEC-developed mobile robot controllers, focusing on those methods that include a physical robot at some point in the learning loop. It simultaneously relates each of these methods to a framework of two orthogonal issues: the use of a simulated and/or a physical robot, and the use of finite, training phase evolution prior to a task and/or lifelong adaptation by evolution during a task. A list of evaluation criteria are presented and each of the surveyed methods are compared to them. Analyses of the framework and evaluation criteria suggest several possibilities; however, there appear to be particular advantages in combining simulated, training phase evolution (TPE) with lifelong adaptation by evolution (LAE) on a physical robot.

Two classes of dynamical recurrent neural

Abstract- A potentially powerful application of evolutionary computation (EC) is to evolve neural networks for automated control tasks. However, in such tasks environments can be unpredictable and fixed control policies may fail when conditions suddenly change. Thus, there is a need to evolve neural networks that can adapt, i.e. change their control policy dynamically as conditions change. In this paper, we examine two methods for evolving neural networks with dynamic policies. The first method evolves recurrent neural networks with fixed connection weights, relying on internal state changes to lead to changes in behavior. The second method evolves local rules that govern connection weight changes. The surprising experimental result is that the former method can be more effective than evolving networks with dynamic weights, calling into question the intuitive notion that networks with dynamic synapses are necessary for evolving solutions to adaptive tasks. 1

Abstract not found

Single-trial learning is studied in an evolved robot model of synaptic spike-timing-dependent plasticity (STDP). Robots must perform positive phototaxis but must learn to perform negative phototaxis in the presence of a short-lived aversive sound stimulus. STDP acts at the millisecond range and depends asymmetrically on the relative timing of pre- and post-synaptic spikes. Although it has been involved in learning models of input prediction, these models require the iterated presentation of the input pattern, and it is hard to see how this mechanism could sustain single-trial learning over a time-scale of tens of seconds. An incremental evolutionary approach is used to answer this question. The evolved robots succeed in learning the appropriate behaviour, but learning does not depend on achieving the right synaptic con¯guration but rather the right pattern of neural activity. Robot performance during positive phototaxis is quite robust to loss of spike-timing information, but in contrast, this loss is catastrophic for learning negative phototaxis where entrained firing is common. Tests show that the final weight configuration carries no information about whether a robot is performing one behaviour or the other. Fixing weights, however, has the effect of degrading performance, thus demonstrating that plasticity is used to sustain the neural activity corresponding both to the normal phototaxis condition and to the learned behaviour. The implications and limitations of this result are discussed.

Plastic spiking neural networks are synthesized for phototactic robots using evolutionary techniques. Synaptic plasticity asymmetrically depends on the precise relative timing between presynaptic and postsynaptic spikes at the millisecond range and on longer-term activity-dependent regulatory scaling. Comparative studies have been carried out for dierent kinds of plastic neural networks with low and high level of neural noise. In all cases, the evolved controllers are highly robust against internal synaptic decay and other perturbations.

Abstract. We give an overview of the EPFL indoor flying project, whose goal is to evolve neural controllers for autonomous, adaptive, indoor micro-flyers. Indoor flight is still a challenge because it requires miniaturization, energy efficiency, and control of non-linear flight dynamics. This ongoing project consists in developing a flying, vision-based micro-robot, a bio-inspired controller composed of adaptive spiking neurons directly mapped into digital micro-controllers, and a method to evolve such a neural controller without human intervention. This document describes the motivation and methodology used to reach our goal as well as the results of a number of preliminary experiments on vision-based wheeled and flying robots. 1.