How do we teach groups of machines autonomy, cooperation and expression?
Nature provides us with many examples of collective behaviors that solve problems and create complexity, community and beauty – from fish, birds and bees to, of course, humans. We can learn from these to create systems of robots, sensors and processes to augment our capabilities and delight us: The future of our society will increasingly be a mesh of human, natural and artificial collectives.
Multi-agent systems that collaborate in physical space to accomplish tasks take many forms, from drone swarms to sensor networks, and their variability and versatility will only increase in the future. Every such system has degrees of autonomy in its control mechanisms: The more autonomous, the more trust is placed in the system’s algorithms to make good decisions, and the less it burdens humans with supervising it.
Artificial Collectives refers to the development of systems that are multitudes of technical units, and how they can be imbued with decentralized, coordinated decision-making. What principles and languages do we need to instruct a swarm of drones to solve a particular problem? How can they communicate with each other, and with their environment? Can we take clues from how groups of living beings cooperate? How can such a system, as a collective, be easily scalable and robust towards imperfect sensing or communication?
Inherent in decentralized agency is a diversity of expressive forms: It may be a philosophical question at what point a decision deserves to be called truly autonomous in a technical apparatus, but we humans certainly perceive autonomy, form and purpose in the coordinated behaviour of a multitude. From this perspective, the expressive potential of a distributed autonomous system is vast: The tension between the artist/programmer, emergent behaviour and perception is here investigated in physical-mechanical space.
This expressive potential shall be used to highlight and discuss the other potentials of this set of technologies: Whether in production, traffic, healthcare or surveillance, existing systems of connected human activity will soon be strongly interwoven with intelligent multi-agent systems, from dynamic automated assembly lines to self-regulating networks of autonomous cars. What, other than increased efficiency, will this mean for society? What, other than supreme convenience, will it add to human experience?
Photos from top left to bottom right: Raphael Schaumburg-Lippe | Gregor Hartl Photography | Martin Hieslmair | NTT
Our technical and conceptual work with Artificial Collectives offers two perspectives for collaborations in different fields:
Expanding the Discourse
Any industry engaged with automating complex activities that mesh deeply with human activity is already facing questions in how robotic and human “collectives” interact. We bring both technical experience and an “Art Thinking”-based approach to the topic; we are experienced in designing open-ended research that expands thought and discussion about such interaction, both among the research partners and with the public.
With swarmOS and a full ecosystem of tools and knowledge for development and operations, we can design a tangible, scalable multi-robot system to explore visions and concepts quickly. Whether the most suitable result is a lab-sized drone/robot interaction scenario or a large-scale public swarm performance: an actual system that can be experienced physically can propel a socio-technological research endeavour forward in ways that a model or simulation cannot.