Many arthropods (particularly insects) exhibit sophisticated visually guided behaviours. Yet in most cases the behaviours are guided by input from a few hundreds or thousands of "pixels " (i.e. ommatidia in the compound eye). Inspired by this observation, we have for several years been exploring the possibilities of visually guided robots with low-bandwidth vision. Rather than design the robot controllers by hand, we use artificial evolution (in the form of an extended genetic algorithm) to automatically generate the architectures for artificial neural networks which generate effective sensory-motor coordination when controlling mobile robots. Analytic techniques drawn from neuroethology and dynamical systems theory allow us to understand how the evolved robot controllers function, and to predict their behaviour in environments other than those used during the evolutionary process. Initial experiments were performed in simulation, but the techniques have now been successfully transferred to work with a variety of real physical robot platforms. This chapter reviews our past work, concentrating on the analysis of evolved controllers, and gives an overview of our current research. We conclude with a discussion of the application of our evolutionary techniques to problems in biological vision.

We present results from the concurrent evolution of visual sensing morphologies and sensory-motor controller-networks for visually guided robots. In this paper we analyse two (of many) networks which result from using incremental evolution with variable-length genotypes. The two networks come from separate populations, evolved using a common fitness function. The observable behaviours of the two robots are very similar, and close to the optimal behaviour. However, the underlying sensing morphologies and sensory-motor controllers are strikingly different. This is a case of convergent evolution at the behavioural level, coupled with divergent evolution at the morphological level. The action of the evolved networks is described. We discuss the process of analysing evolved artificial networks, a process which bears many similarities to analysing biological nervous systems in the field of neuroethology.

This paper has not already been accepted by and is not currently under review for a journal or another conference. Nor will it be submitted for such during the IJCAI's review period

Although artificial and biological systems face similar sensorimotor control problems, until today only a few attempts have been made to implement specific biological control structures on robots. Nevertheless, the process of designing the sensorimotor control of a robot can contribute to our understanding of these mechanisms and can provide the basis of a critical evaluation of existing biological models. Flies have developed a specialized visuomotor control for tasks such as course stabilization, fixation and approach towards stationary objects, tracking of moving objects and landing, which are based on the analysis of visual motion information. Theoretical and experimental results suggest that in flies the visuomotor control for course stabilization as well as fixation and approach towards stationary objects may be implemented at least partially by one common sensory circuit. We present agents with a visuomotor controller that regulates the two behaviors of course stabilization and object fixation. To test this controller under real world conditions, we implemented it on a miniature robot. We have been able to show that in addition to course stabilization and object fixation, the robot also approaches stationary

. This article describes a 3D biomechanical simulation of a salamander to be used in experiments in computational neuroethology. The physically-based simulation represents the salamander as an articulated body, actuated by muscles simulated as springs and dampers, in interaction with a simple environment. The aim of the simulation is to investigate the neural circuits underlying the aquatic and terrestrial locomotion of the real salamander, as well as to serve as test bed for investigating vertebrate sensorimotor coordination in silico. 1 Computational neuroethology This article describes the design of a 3D biomechanical model for experiments in computational neuroethology [1, 2], a field which studies how behaviors of an autonomous agent (a simulated animal or a robot) arise from neural circuits. A central aspect of computational neuroethology is that it integrates the simulated central nervous system into a body and an environment, and that it investigates behavior as the re...

. This paper briefly reviews synthetic approaches to neurobiology and presents results of two experiments on the use of evolutionary algorithms for the design of neural controllers for locomotion. The first experiment consists in using the evolutionary algorithm for instantiating low level parameters of a connectionist simulation of the lamprey's locomotor circuitry. The second experiment develops potential neural circuits for the swimming and trotting of the salamander; an animal whose locomotor circuitry has currently not been decoded. In both cases, biologically plausible control circuits are developed which produce a neural activity with many similarities to that measured in the real animals. 1 Synthetic approaches to neurobiology The fields of artificial life and artificial intelligence have developed tools and methods which have the potential to significantly help computational neurobiology. Synthetic approaches to neurobiology can indeed increase our understanding of ...

This article presents a first step towards our aim of building comprehensive models of neural systems embedded into biomechanical simulations for investigating the neuroethology of lower vertebrates. It investigates visual tracking in a model of the salamander which incorporates 1) a biomechanical body actuated by springs-and-damper muscles, 2) a spinal locomotor circuit simulated as a leaky-integrator neural network, and 3) a visionbased control circuit. A simple visuomotor experiment is presented in which the vision-based control circuit uses the signals from the retinas and from "water" sensors to modulate the type of gait produced by the locomotor circuit as well as the direction of motion. The simulated salamander is thus given the capacity to track and approach a randomly moving target both in water and on ground. Although our aim is to integrate realistic models of the visual system into the salamander model, only a very abstract visual system is implemented ...

There are striking parallels between ecological psychology and the new trends in robotics and computer vision, particularly regarding how agents interact with the environment. We present some ideas from ecological psychology, including control laws using optic flow, affordances and action modes, and describe our implementation of these concepts in a small mobile robot which can avoid obstacles and chase or flee moving targets solely using optic flow. The properties of these methods were further explored in simulation. This work ties in with that of others arguing for a methodological approach in robotics which foregoes a central model/planner. Ecological psychology may not only contribute to robotics, but robotic implementations may in turn provide a test bed for ecological principles and a source of ideas which could be tested in animals and humans. Keywords: ecological psychology, behavior-based robotics, optic flow, obstacle avoidance, tag. ECOLOGICAL ROBOTICS 2 Introduction Over...

This paper discusses the relationship between Artificial Intelligence (AI) and Artificial Life (A-Life). A-Life research addresses a wide range of phenomena, some of which have no obvious bearing on AI research. The work most relevant to AI is sufficiently coherent and distinct that it is best referred to by its own name: it is Adaptive Behavior research which is most likely to have significant impact on issues traditionally studied in AI. Some motivations for adaptive behavior research are reviewed, and some of the differences between adaptive behavior and traditional AI are discussed. One significant feature of current adaptive behavior research is a focus on relatively simple and specialised cognitive functions, an approach which invites unfavourable comparisons with the "blocksworld" simplified domains which were popular in AI research of the early 1970's. However, such comparisons usually overlook fundamental differences between the blocksworld-AI and Adaptive Behavior approaches ...

The study of biological systems has inspired the development of a large number of neural network architectures and robotic implementations. Through both experimentation and simulation biological systems provides a means to understand the underlying mechanisms in living organisms while inspiring the development of robotic applications. Experimentation, in the form of data gathering (ethological physiological and anatomical), provides the underlying data for simulation generating predictions to be validated by theoretical models. These models provide the understanding for the underlying neural dynamics, and serve as basis for simulation and robotic experimentation. Due to the inherent complexity of these systems, a multi-level analysis approach is required where biological, theoretic and robotic systems are studied at different levels of granularity. The work presented here overviews our existing modeling approach and describes current simulation results. 1 Introduction The st...