Research Area "Neurocybernetics"

 
Home
Research topics
Projects
Tools
Publications
People
Links
Contact

Doctorate Programme
Institute of Cognitive Science
University of Osnabrück

Octavio - Neurocybernetics of a multilegged walking machine

Octavio project overview video (MP4, WebM, or Ogg).


The Octavio project relies on a tight interaction between (neuro-)biology and robotics to gain new insights into the principles underlying adaptive and multifunctional control of walking in biology as well as in robotics. It is based upon the concept of cybernetically equivalent processes in biology and technology. The project consists of four parts: 1. the physical walking machine Octavio, 2. neural control consisting of evolved neural networks and/or functional cybernetic models, derived from the (neuro-)biology of the stick insect, 3. biological-experiments, and 4. embedding results in a neurocybernetics theory.


The walking machine Octavio was designed as a modular machine in the sense that its legs constitute autonomous units which may be attached to diversly shaped torsi in an arbitrary number. Each leg has sensori-motor control over its three joints, multiple exo- and proprioceptors as well as its own energy supply. Legs attached to a common body communicate via a simple bus-system. High-level control systems that do not participate in the mere walking and which might use additional external sensors. e.g. cameras, are placed inside the torso.


Initially -- by using a physical simulation of the realised walking machine with given morphology (e.g. 4-, 6-, 8-legged) -- evolutionary methods are employed to develop neuro-control structures for single legs. Suitable neuro-controllers are then implemented and tested on the physical machine. In parallel functional cybernetic models of the nervous system of the stick insect are evaluated in simulation and on the physical machine.


By comparing functional models for locomotion control deduced from the living organisms with those derived from evolution of artificial neuro-controllers the following benefits are expected: 1. development of a highly-efficient walking machine and corresponding neural control, 2. new basic principles of neural control of locomotion in the sensori-motor loop, 3. hypotheses that lead to new experiments in (neuro-)biology and 4. an embedding in the context of neuro-cybernetical theory of the nervous system.





Contact: Frank Pasemann and Arndt von Twickel

Project partners: Animal Physiology lab in Cologne (Prof. Dr. Ansgar Büschges).

Supported by