Similar to the Autonomous Computing initiative, which has mainly been advancing techniques for self-optimization focusing on computing systems and infrastructures, Organic Computing (OC) has been driving the development of system design concepts and algorithms for self-adaptive systems at large. Examples of application domains include, for instance, traffic management and control, cloud services, communication protocols, and robotic systems. Such an OC system typically consists of a potentially large set of autonomous and self-managed entities, where each entity acts with a local decision horizon. By means of cooperation of the individual entities, the behavior of the entire ensemble system is derived. In this article, we present our work on how autonomous, adaptive robot ensembles can benefit from OC technology. Our elaborations are aligned with the different layers of an observer/controller framework, which provides the foundation for the individuals’ adaptivity at system design-level. Relying on an extended Learning Classifier System (XCS) in combination with adequate simulation techniques, this basic system design empowers robot individuals to improve their individual and collaborative performances, e.g., by means of adapting to changing goals and conditions. Not only for the sake of generalizability but also because of its enormous transformative potential, we stage our research in the domain of robot ensembles that are typically comprised of several quad-rotors and that organize themselves to fulfill spatial tasks such as maintenance of building facades or the collaborative search for mobile targets. Our elaborations detail the architectural concept, provide examples of individual self-optimization as well as of the optimization of collaborative efforts, and we show how the user can control the ensembles at multiple levels of abstraction. We conclude with a summary of our approach and an outlook on possible future steps.
An Organic Computing Approach to Self-Organizing Robot Ensembles
Sebastian von Mammen, Sven Tomforde and Jörg Hähner
Front. Robot. AI, 17 November 2016 | https://doi.org/10.3389/frobt.2016.00067