Justin Werfel

Collective construction
RoboBees
Amoebots
Other swarm and bioinspired robotics
Engineered morphogenesis
Evolutionary theory
DNA self-assembly
Cancer modeling
Computational neuroscience
Other fun stuff

I'm a research scientist at Harvard's Wyss Institute for Biologically Inspired Engineering, where I work on topics in complex and emergent systems, with a focus on swarm robotics. I work most closely with the Self-Organizing Systems Research Group, and interact also with the Molecular Systems Lab and Microrobotics Lab.

Other fun stuff
Computational neuroscience
Cancer modeling
DNA self-assembly
Evolutionary theory
Engineered morphogenesis
Other swarm and bioinspired robotics
Amoebots
RoboBees
Collective construction



Research - Personal - CV



I'm interested in understanding and designing complex systems—systems of many independent interacting components, where we may understand each component very well in isolation, but when many of them get together, some interesting new high-level behavior emerges. Can we predict that collective outcome from the rules the low-level agents follow? And can we design low-level behaviors that guarantee a particular high-level result? I've looked at these questions in topic areas including swarm robotics, evolutionary theory, and DNA self-assembly.


Collective construction

Termite colonies build tremendous, complicated mounds, acting with no central control or careful advance planning. These social insects provide a fantastic proof of principle that (relatively) simple agents, acting independently with access only to local information, can build amazing things. How could we build and program robot swarms—artificial termite colonies—to build things for us? We want a human user to be able to give such a swarm a high-level description of what they want built, and have a guarantee that the system will build that thing, without the user having to get into the details of how it's done.

Read more about earlier work or the more recent TERMES project, which I co-lead with Radhika Nagpal.

Media coverage (To be updated!!): Discover Magazine (April 2013), Communications of the ACM (March 2013), Reuters TV (June 2012), IEEE Spectrum (Automaton blog, June 2011), Boston Globe (March 2010), CNET News (July 2006), Wired (July 2006).

Publications:

Designing collective behavior in a termite-inspired robot construction team. Justin Werfel, Kirstin Petersen, and Radhika Nagpal. Science 343(6172): 754-758 (2014).
[For free access without a subscription, please follow the links here.]

TERMES: an autonomous robotic system for three-dimensional collective construction. Kirstin Petersen, Radhika Nagpal, and Justin Werfel. Robotics: Science and Systems VII, pp. 257-264 (2011).

Distributed multi-robot algorithms for the TERMES 3D collective construction system. Justin Werfel, Kirstin Petersen, and Radhika Nagpal. Workshop on Reconfigurable Modular Robotics, at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2011.

Collective construction with robot swarms. Justin Werfel. In Morphogenetic Engineering, Rene Doursat, Hiroki Sayama, and Olivier Michel, eds., Springer, pp. 115-140 (2012).

Three-dimensional construction with mobile robots and modular blocks. Justin Werfel and Radhika Nagpal. International Journal of Robotics Research 27 (3-4): 463-479 (2008).

Collective construction of environmentally-adaptive structures. Justin Werfel, Donald Ingber, and Radhika Nagpal. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2007.

Towards a common comparison framework for global-to-local programming of self-assembling robotic systems. Justin Werfel and Radhika Nagpal. Workshop on Self-Reconfigurable Robots & Systems and Applications, at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2007.
Full version: Technical Report TR-14-07, Harvard EECS, 2007.

Robot search in 3D swarm construction. Justin Werfel. First IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO), pp. 363-366 (2007).

Anthills built to order: Automating construction with artificial swarms. Justin Werfel. Doctoral thesis, MIT, May 2006.

Extended stigmergy in collective construction. Justin Werfel and Radhika Nagpal. IEEE Intelligent Systems 21(2): 20-28 (2006).

Distributed construction by mobile robots with enhanced building blocks. Justin Werfel, Yaneer Bar-Yam, Daniela Rus, and Radhika Nagpal. IEEE International Conference on Robotics and Automation (ICRA), pp. 2787-2794 (2006).

Three-dimensional directed construction. Justin Werfel and Radhika Nagpal. Workshop on Self-Reconfigurable Modular Robots, at Robotics: Science and Systems II, 2006.

Collective construction using LEGO robots. Crystal Schuil, Matthew Valente, Justin Werfel, and Radhika Nagpal. Robot Exhibition, Twenty-First National Conference on Artificial Intelligence (AAAI), 2006. [Received Technical Innovation Award for "elegant connection of theory and design".]

Building patterned structures with robot swarms. Justin Werfel, Yaneer Bar-Yam, and Radhika Nagpal. Nineteenth International Joint Conference on Artificial Intelligence (IJCAI), pp.1495-1502 (2005).

Construction by robot swarms using extended stigmergy. Justin Werfel, Yaneer Bar-Yam, and Radhika Nagpal. AI Memo AIM-2005-011, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), 2005.

Building blocks for multi-agent construction. Justin Werfel. Distributed Autonomous Robotic Systems 6 (DARS), 2004.



RoboBees

Inspired by honeybees, the purpose of the Micro Air Vehicles project is to develop hardware and algorithms for swarms of thousands of tiny, self-contained flying robots, to perform tasks such as commercial pollination. I'm involved with the "Colony" part of this project, investigating coordination mechanisms for autonomous robots with extremely limited capabilities.

Publications:

Positional communication and private information in honeybee foraging models. Peter Bailis, Radhika Nagpal, and Justin Werfel. International Conference on Swarm Intelligence (ANTS), 2010. (Best Student Paper Award)



Slime mold robots (Amoebots)

Cellular slime molds have a life cycle in which millions of individual amoebae stream together into multicellular slugs, which then crawl as a single coordinated unit. We're interested in creating a robot version, partly because it would be just about the coolest thing ever but also because such a robot could be used for things like extraplanetary exploration, with individual modules exploring on their own for fast parallel coverage or coming together for improved mobility to overcome obstacles.

Publications:

Coordinating collective locomotion in an amorphous modular robot. Chih-Han Yu, Justin Werfel, and Radhika Nagpal. IEEE International Conference on Robotics and Automation (ICRA), 2010.



Other swarm and bioinspired robotics

Publications:

Massive uniform manipulation: controlling large populations of simple robots with a common input signal. Aaron Becker, Golnaz Habibi, Justin Werfel, Michael Rubenstein, and James McLurkin. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2013.

Collective transport of complex objects by simple robots: theory and experiments. Michael Rubenstein, Adrian Cabrera, Justin Werfel, Golnaz Habibi, James McLurkin, and Radhika Nagpal. International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), 2013.

Collective decision-making in multi-agent systems by implicit leadership. Chih-Han Yu, Justin Werfel, and Radhika Nagpal. International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), 2010.

Bioinspired environmental coordination in spatial computing systems. Justin Werfel, Yaneer Bar-Yam, and Donald Ingber. Workshop on Spatial Computing, at Second IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO), 2008.

Research, robots, and reality: a statement on current trends in biorobotics. Ernst Niebur et al. Behavioral and Brain Sciences 24(6):1072-1073 (2001).



Engineered morphogenesis

Morphogenesis, the process of development from a single cell to a complex multicellular organism, is one of the great examples of robust collective behavior. How could we specify genetic programs to grow plants and animals in specific forms or body plans we want?

Publications:

Biologically realistic primitives for engineered morphogenesis. Justin Werfel. International Conference on Swarm Intelligence (ANTS), 2010.



Evolutionary theory

If evolution is driven by competition, and an organism's main competitors are others of its same species, why should cooperation be as widespread and successful as it evidently is? Modeling studies have helped elucidate mechanisms behind the evolution of altruistic behaviors, relevant to issues including multilevel selection, intraspecific communication and sociality.

Other work looks at questions relevant to evolutionary algorithms and engineering applications. How can coevolving populations avoid "Red Queen" arms races with no net progress against an external metric? Can evolving populations be harnessed to perform useful computations?

Publications:

Multilevel and kin selection in a connected world. M.J. Wade, D.S. Wilson, C. Goodnight, D. Taylor, Y. Bar-Yam, M.A.M. de Aguiar, B. Stacey, J. Werfel, G.A. Hoelzer, E.D. Brodie III, P. Fields, F. Breden, T.A. Linksvayer, J.A. Fletcher, P.J. Richerson, J.D. Bever, J.D. Van Dyken, and P. Zee. Nature 463: E8-E9 (2010).

The evolution of reproductive restraint through social communication. Justin Werfel and Yaneer Bar-Yam. Proceedings of the National Academy of Sciences (PNAS) 101(30): 11019-11024 (2004).

Modeling, communication, and global catastrophe. Justin Werfel and Yaneer Bar-Yam. Knowledge Magazine 1(1):5-13 (2009).

Resource sharing and coevolution in evolving cellular automata. Justin Werfel, Melanie Mitchell, and James P. Crutchfield. IEEE Transactions on Evolutionary Computation 4:388-393 (2000).

Implementing universal computation in an evolutionary system. Justin Werfel. AI Memo AIM-2002-010, MIT Artificial Intelligence Lab, 2002.



DNA self-assembly

The specificity with which DNA bases pair together makes it interesting not only as an informational molecule but also as a structural one. How can we design DNA sequences so that strands self-assemble into structures, or dynamic processes involving binding and unbinding, and how can we automate elements of the design process?



Cancer modeling

A tissue is a complex system of cells interacting: exchanging signals, exerting physical forces on one another, and otherwise shaping their environment. Computational models let us look at both cell-level and tissue-level phenomena, and explore possibilities like how cell interactions could evoke or reverse a cancerous phenotype even without genetic mutation.

Media coverage: Biomedical Picture of the Day.

Publications:

How changes in extracellular matrix mechanics and gene expression variability might combine to drive cancer progression. Justin Werfel, Silva Krause, Ashley G. Bischof, Robert J. Mannix, Heather Tobin, Yaneer Bar-Yam, Robert M. Bellin, and Donald E. Ingber. PLOS ONE 8(10): e76122 (2013).



Computational neuroscience

The brain is another canonical complex system: collections of billions of neurons give rise to the nearly magical phenomenon of consciousness. Are there learning processes we can formally analyze, that could both realistically be taking place in biological cells and operate on relevant time scales? Can we extract signals from EEG recordings that are reliably correlated with intent, and could be used as a controller for a prosthesis or communication device? What mechanisms of synaptic transmission are necessary to produce certain observed neuronal firing patterns?

Publications:

Learning curves for stochastic gradient descent in linear feedforward networks. Justin Werfel, Xiaohui Xie, and H. Sebastian Seung. Neural Computation 17(12): 2699-2718 (2005).

BCI Competition 2003--data set Ia: combining gamma-band power with slow cortical potentials to improve single-trial classification of electroencephalographic signals. Brett Mensh, Justin Werfel, and H. Sebastian Seung. IEEE Transactions on Biomedical Engineering 51(6):1052-1056 (2004).

Learning curves for stochastic gradient descent in linear feedforward networks. Justin Werfel, Xiaohui Xie, and H. Sebastian Seung. Neural Information Processing Systems 16 (NIPS), 2004.

Neural network models for zebra finch song production and reinforcement learning. Justin Werfel. Master's thesis, MIT, August 2001.




I'm a big proponent of the idea that the things one does outside work are a major part of what makes life interesting. For me, these have included activities like—

performing arts: juggling, storytelling [e.g., The Moth (StorySLAM Winner), MassMouth (season finalist 2011, 2012; Audience Choice award at 2011 finals), Story Collider], a cappella and light opera, musical theater, pantomime;

visual arts: glassblowing, blacksmithing;

writing (Active member of the SFWA);

solo sports: skiing, diving, long-distance and mountain unicycling.