Emerging computing technologies have made it possible to manufacture large-scale embedded multi-agent systems, from vast sensor networks to modular robots and smart materials. A key challenge is understanding how to program such systems at the individual agent level in order to achieve system-level behavior that is complex, fault-tolerant, and adapts to the environment. One source of inspiration is multicellular biological systems (tissues, organs, and simple organisms) that achieve complex self-adaptation in changing environments through the distributed cooperation and sensing of vast numbers of cells. Such systems can provide novel bio-inspired principles for the design and programming of multi-agent systems.
This research investigates new computational paradigms for programming multi-agent robotic systems to achieve complex self-adaptation in response to the environment. The two main thrusts are: (a) the development of a global-to-local programming methodology for describing complex global adaptation goals and automatically deriving provably robust multi-agent control (b) the development of a tissue-inspired modular robotic system that can demonstrate self-adaptive structures. A key source of inspiration is the decentralized control strategies that cells use to achieve environment-responsive structures and functions, e.g. shape adaptation in plants and vascular networks, and locomotion in simple animals. The aim is to harness these bio-inspired principles to create multi-agent robotic systems that replicate the adaptability and responsiveness of living systems. This work has broad application, from robust control in distributed sensor-actuator networks to the development of self-adapting architectural structures and prosthetics. This research significantly advances our understanding of how to design autonomous multi-agent computing systems that can self-organize, self-repair, and respond to the environment.
