Advances in technology have begun to allow for the production of large groups, or swarms, of robots; however, there exists a large gap between their current capabilities and those of swarms found in nature or envisioned for future robot swarms.  These deficiencies are the result of two factors, difficulties in algorithmic control of these swarms, and limitations in hardware capabilities of the individuals.  Creating a hardware system for large robotic swarms is an open challenge; cost and manufacturability pressure hardware designs to be simple with minimal capabilities, while algorithm design favors more capable hardware. The robot design must balance these factors to create a simple robot that is, at the same time, capable of performing the desired behaviors.  To investigate these challenges, I created the Kilobot robot swarm, a swarm of 1024 (“kilo”) robots.  In this talk, I will discuss the many challenges associated with creating a robot swarm at this scale and the implications this has for creating even larger, more capable swarms in the future.

Controlling these swarms is also a challenge, as the properties desired from these systems, e.g. shape, locomotion, are generally a global property; however, we can only control local interactions between individuals. Furthermore, the mapping between controllable local behaviors and desired global results is not well understood. Their control is further complicated by the very nature of these systems which are composed of decentralized, distributed, asynchronous, error-prone individuals with often limited capabilities.  I will discuss two examples of algorithms recently implemented on the Kilobot swarm, self-assembly of user-defined 2D shapes, and the collective transport of objects. Both of these examples provide guarantees of correctness and performance bounds of the swarm, and provide examples of reliable global-to-local control over a robot swarm.  I will describe unexpected challenges faced while trying to control the Kilobot swarm, and how these challenges will influence the design of future swarm algorithms.

Michael Rubenstein is currently a researcher in Radhika Nagpal's Self-Organizing Systems Research Group at Harvard University's School of Engineering and Applied Sciences.  There he is working on Kilobot, a robot designed for testing swarm algorithms in a group of over a thousand robots.  He received his Ph.D. from The University of Southern California's School of Computer Science under the supervision of Wei-Min Shen.  His thesis, titled: "Self-Assembly and Self-Healing for Robotic Collectives", details a control algorithm for a simple, simulated multi-robot system which guarantees that it can self-assemble and self-heal any desired connected shape.  Most of his research is centered around the design and control of multi-robot systems.  Additional information can be found at his webpage: http://people.seas.harvard.edu/~mrubenst/