In nature, social insects have shown how mobility and multiplicity can be dramatically exploited: small ants effectively forage large areas, while tiny termites construct structures a million times their own size. Yet engineering a robotic swarm of similar complexity remains a significant challenge; we lack a fundamental understanding of how to effectively implement and leverage swarm behaviors. In this research we will design and implement a programmable microrobot swarm capable of robust collective intelligence in complex environments. Our project will focus on mobility and multiplicity: we seek to understand how scaling the agent size down and the swarm size up impacts the ability of collective networked systems. Our proposed work has two key threads: 1) the development of highly mobile insect-scale microrobots with biologically-derived locomotion principles and 2) the design and mathematical analysis of robust decentralized control paradigms which are broadly applicable to many self-organizing complex networks. To validate our approach, we will integrate the mechanical study of mobility and the algorithmic study of collectives through a set of annual demonstrations of microrobot swarms, which will enable continuous feedback between theory and engineering.
In nature, social insects have shown how mobility and multiplicity can be dramatically exploited: small ants effectively forage large areas, while tiny termites construct structures a million times their own size. In the past, engineering a robotic swarm of similar complexity was a significant challenge due to a lack of fundamental understanding of how to effectively implement and leverage swarm behaviors. This project has explored a variety of algorithms for multi-robot coordination, with an emphasis on implementing complex behaviors using simple robots. We have developed foraging algorithms and implemented them on a large-scale, fully autonomous robotic swarm, called the "Kilobot swarm". This integration has allowed us to test the scalability of ant-inspired algorithmic behavior on actual robots, with hundreds to thousands of centimeter-scale robots. In parallel, we have created two classes of terrestrial microrobots. These robots serve to explore the locomotion capabilities of small robots and inform the design of coordination algorithms that rely on this information. Specifically, we have achieved a fully autonomous 1.7g hexapod robot and created a multi-segmented centipede-inspired robot. The former is currently the smallest fully-autonomous hexapod robot and the latter serves as a platform for testing locomotion performance as a function of multiple gait patterns. These two robots have leveraged recent advances in microfabrication processes developed at Harvard. Together with the work from our other research efforts, this project has addressed algorithmic, design, fabrication and validation questions related to operating swarms of terrestrial microrobots. We have also leveraged the results of this project to enhance our outreach and education activities. Robot prototypes and demonstrations have enhanced many of our K-12 lectures and outreach materials.
