Growing & Moving Reconfigurable Robot Organisms

In the project SYMBRION robots are "grown" from having individually moving cells aggregating in specific configurations and then physically docking to each other. This way a larger "robotic organism" can grow, and also reconfigure. Every individual robotic module in this organisms acts autonomously, just like a cell in your body. It can communicate with its neighboring cells only and it has actuators to deform itself with an internal hinge. This way the whole robotic organism can start to move, e.g., by crawling, but only if each individual robotic cell coordinates its local actions well with the others. Bio-inspired algorithms, taken from embryogenesis shape the organism formation process, while algorithms inspired from hormone interaction networks and from neural networks take over the control of the locomotion. These software systems allow evolutionary computation algorithms to find their own way to move the specific body, and nobody, no human programmer or operator, has to sit down and program behavioral programs for these robots.

A Simple Robot Organism Learns To Move on its Own

Three robot modules coordinate their actions in order to effectively move the body forward. This requires already quite complex coordination between these three autonomous robotic "cells" that compose this small "organism". The robots can only interact with their neighboring modules, to keep the system scalable. The motion sensing is fed back to the central module which releases an emulated "success" hormone, which informs all other modules via a simulated diffusion process in the body.

A Larger Organism Can Learn it Too

A large "organism" built from 7 autonomous modules teaches itself how to move by bending its 3 arms in a coordinated way. This robot does not only require more communication because it has more modules, it has also more than a doubled weight compared to his smaller mate on the left. Thus, moving forwards is even more difficult, requiring even more efficient motion gaits.

Evolve it!
Then move it!

In computer simulations I apply a combination of bio-inspired algorithms (evolutionary, hormonal, self-organizational) to find good gaits. Colors show the local and dynamic virtual hormone levels as they flow though the body cell by cell.

Some Typical Symbrion Robots

The following section shows a collection of robotic organisms form the Symbrion project

My research in the EU FET project SYMBRION is in cooperation with the following international and interdisciplinary partners: Universität Stuttgart, Germany; Universität Karlsruhe, Germany; Universität Tübingen, Germany; Vlaams Instituut voor Biotechnologie; Belgium; Czech Technical University in Prague, Czech Republic; Universite Libre de Bruxelles, Belgium; Institut National de Recherche Eninformatique et Automatique, France; Vrije Universiteit Amsterdam, Netherland, Centre National de la Recherche Scientifique, Paris, France; University of York, United Kingdom. In my lab, Dr. Ronald Thenius, Dr. Payam Zahadat, Mag. Christoph Möslinger conduct significant work in this project.