Engineering Self-Organizing Multi-Agent Systems

In this demo session, we will present two examples of how one can systematically program self-organizing multi-agent systems, using inspiration from biology. The first system is a modular robot that autonomously adapts to satisfy complex environmentally-adaptive goals through the cooperation of multiple module agents. The second is a global-to-local compiler that can transform a user-specified pattern formation goal into a multi-agent program and reason about the agent resources required. These systems show (1) how biological design principles can be formally captured and theoretically analyzed, and (2) how global goals can be translated into local interactions amongst many simple agents. Both systems will be demonstrated in real-time and interactively with the audience.

Authors

• Radhika Nagpal Harvard University
• Chih-Han Yu Harvard University
• Daniel Yamins

Keywords

• Self-organization
• global-to-local compilers
• cellular automata