How to Engineer Robotic Organisms and Swarms? - Bio-Inspiration, Bio-Mimicry, and Artificial Evolution in Embodied Self-Organized Systems

In large-scale systems composed of autonomous embodied agents (e.g., robots), unpredictability of events, sensor noise and actuator imperfection pose significant challanges to the designers of control software. If such systems tend to selforganize, emergent phenomena prevent classical engineering approaches per se. In recent years, the Artificial Life Lab at the University of Graz has investigated a variety of methods to synthesize such control algorithms used in multi-modular robotics and in swarm robotics. These methods either translate mechanisms directly from biology to the engineering domain (bio-mimicry, bio-inspiration) or generates such controllers through artificial evolution from scratch. In this article I first discuss distributed control algorithms, which determine the collective behavior of autonomous robotic swarms. These algorithms are derived from collective behavior of honeybees and from slime mold aggregation. One of these algorithms is inspired by inter-adult food exchange in honeybees (’trophallaxis’) another one from chemical signaling in slime molds. In addition to the control of robot swarms, control paradigms for multi-modular robotic organisms are presented, which are again based on simulated fluid exchange (hormones) among compartments of robotic organisms. In both domains -swarms and organisms- the control system is self-organized and consists of many homeostatic sub-systems which adapt to each other on the individual (module) and on the collective level (organism, swarm). Additionally, I discuss the importance of distributed feedback networks, as well as the benefits and drawbacks of bio-inspiration and bio-mimicry in collective robotics.