Spatial Temporal Patterns for Action-Oriented Perception in Roving Robots

Book Preface

The basic principles guiding sensing, perception and action in bio systems seem to rely on highly organised spatial-temporal dynamics. In fact, all biological senses, (visual, hearing, tactile, etc.) process signals coming from different parts distributed in space and also show a complex time evolution. As an example, mammalian retina performs a parallel representation of the visual world embodied into layers, each of which represents a particular detail of the scene. These results clearly state that visual perception starts at the level of the retina, and is not related uniquely to the higher brain centres. Although vision remains the most useful sense guiding usual actions, the other senses, first of all hearing but also touch, become essential particularly in cluttered conditions, where visual percepts are somehow obscured by environment conditions. Efficient use of hearing can be learnt from acoustic perception in animals/insects, like crickets, that use this ancient sense more than all the others, to perform a vital function, like mating. Also motion in living systems is derived from highly organised neural networks driving limbs and other parts of the body, in response to sensory inputs. A huge number of muscular cells are controlled in real time and in an adaptive fashion. Biological locomotion generation and control patterns were demonstrated to be efficiently modelled by using the paradigm of the Central Pattern Generator (CPG) as well as the decentralised control approach (Walknet). In locomotion, different and differently specialised neuronal assemblies behave in a self-organised fashion in such a way that, as a consequence of certain stimuli, particular patterns of neural activity arise, which are adeguately sent to peripheral fibres to generate rhythmic activities of leg motion and control. Recent results have also shown that nonlinear and complex dynamics in cellular circuits and systemscan efficiently model both sensing and locomotion. After analysing and being involved in such topics, researchers from various scientific fields and from different European Laboratories came up to the idea to glue their effort to try to go up from sensing and locomotion modelling, to perception. The present volume collects the most significant results of the European research Project called “SPARK”, whose aim was to develop completely new sensing-perceiving-moving artefacts inspired by the basic principles of living systems and based on the concept of “self-organization”. The project activities were carried out by scientists from six institutions: the Institute for Biological Cybernetics of the University of Bielefeld (Germany), the Institute of Perception, Action and Behaviour, University of Edinburgh (UK), the Instituto Pluridisciplinar, Universidad Complutense deMadrid (Spain), the Companies ANALOGIC s.l. from Budapest (Hungary) and Innovaciones Microelectronicas s.l., ANAFOCUS, from Seville (Spain). The Consortium was Coordinated by the Dipartimento di Ingegneria Elettrica Elettronica e dei Sistemi of the Unversity of Catania (Italy).

Contents

Part I: Systems

1 Perception for Action in Insects .

2 Principles of Insect Locomotion

3 Low Level Approaches to Cognitive Control

Part II: Cognitive Models

4 A Bottom-Up Approach for Cognitive Control .

5 Mathematical Approach to Sensory Motor Control and Memory .

6 From Low to High Level Approach to Cognitive Control

7 Complex Systems and Perception

Part III: Software/Hardware Cognitive Architecture and Experiments

8 New Visual Sensors and Processors .

9 Visual Algorithms for Cognition

10 SPARK Hardware

11 Robotic Platforms and Experiments