Skip to main content

Control and Optimization for Robot Swarms


A wide range of networked systems exhibit emergent behavior. In nature, for example, flocks of birds, schools of fish, and swarms of bees all develop cohesive global behavior from purely local interactions. The goal of our research is to develop the tools necessary to design local control, communication, and estimation laws for individual agents that yield a desired group behavior. For example, we might want to design controllers to make the agents achieve a desired formation, surround an enemy agent, or move to efficiently measure a quantity in their environment. 

A major thrust of our work is to develop a theoretical design framework for coordinating multi-agent systems. Our primary design approach is a multi step process. First we develop a centralized controller that will make the swarm perform the desired task. This controller can rely on global information about the swarm, such as its center of mass. We derive a decentralized version of the initial controller by replacing the global information with the output of a local estimator running on each agent. The estimator combines information received from the agent's neighbors with that agent's own sensor readings to approximate the global quantity.

In addition to our theoretical work, we also test our theory using a swarm of "e-puck" mobile robots designed by École Polytechnique Federale de Lausanne. Each e-puck is modified to include an XBee radio for communication and a color sensor. The color sensor measures the light coming from a projector that displays images over the arena. The intensity of this light simulates an environmental quantity of interest such as temperature or chemical concentration, and the color sensor allows the robots to interact with this environment. Using an earlier version of this setup that did not include the color sensor or projector, we demonstrated simple formation control. With the addition of the projector and color sensor, we can now experiment with more complex tasks such as environmental modeling.


Kevin M. Lynch
Randy Freeman
Matthew Elwin
Samhitha Poonacha
Andrew Kessler


Related Publications

M. Hwang, M. L. Elwin, P. Yang, R. A. Freeman, and K. M. Lynch, Experimental Validation of Multi-Agent Coordination by Decentralized Estimation and Control, Networking Humans, Robots, and Environments, Bentham Science, 2013

H. Bai, R. A. Freeman, and K. M. Lynch, Distributed Kalman Filtering Using the Internal Model Average Consensus Estimator, American Control Conference (ACC), 2011, IEEE, pp. 1500-1505, 2011 Google Scholar

F. Morbidi, R. A. Freeman, and K. M. Lynch, Estimation and control of UAV swarms for distributed monitoring tasks, American Control Conference (ACC), 2011, IEEE, pp. 1069-1075, 2011 Google Scholar

H. Bai, R. A. Freeman, and K. M. Lynch, Robust dynamic average consensus of time-varying inputs, Decision and Control (CDC), 2010 49th IEEE Conference on, IEEE, pp. 3104-3109, 2010 Google Scholar

R. A. Freeman, T. R. Nelson, and K. M. Lynch, A complete characterization of a class of robust linear average consensus protocols, American Control Conference (ACC), 2010, IEEE, pp. 3198-3203, 2010 Google Scholar

P. Yang, R. A. Freeman, G. J. Gordon, K. M. Lynch, S. S. Srinivasa, and R. Sukthankar, Decentralized estimation and control of graph connectivity for mobile sensor networks, Automatica, vol. 46, Elsevier, pp. 390-396, 2010 Google Scholar

Back to top