home · mobile · calendar · colloquia · 2008-2009 · 

Colloquium - Correll

Modeling and Design of Large-Scale Sensor-Actuator Networks
Massachusetts Institute of Technology

Networks of sensors and actuators have the potential to provide intelligent and robust solutions for a variety of real-world sensing and manipulation problems. Mobility, sensing, and actuation in such systems are not limited to artificial agents, but involve plants, animals, and people. Instances of such systems that I will present are robot swarms for inspection of machinery, in-car devices for gathering real time traffic information, algorithms for herding cattle using smart collars, mobile ad-hoc wifi networks, a distributed robotic garden that takes care of tomato plants and a prototype of a "smart material" consisting of networked computationally enhanced actuators. Key challenges when designing and modeling such systems are (a) how to optimally distribute sensing, actuation and computation in the environment and (b) how to predict the performance of a specific distributed system subject to varying assumptions on hardware performance and features.

I will illustrate these challenges focusing on a multi-robot inspection case study, for which I developed a series of algorithms ranging from purely reactive to fully deliberative coordination, each with increasing performance and hardware requirements. I will then outline a general methodology for developing probabilistic models of such systems that take the uncertainty of the agents' actions into account and allow for formally studying resource trade-offs in the system. I will demonstrate the viability of this approach using results from real robot experiments with swarms of up to 40 miniature robots, and outline how the methodology applies to other systems that I am currently working with.

Nikolaus Correll is a post-doctoral associate at the Computer Science and Artificial Intelligence Laboratory, MIT, where he works with Daniela Rus. Nikolaus research interests focus on networked intelligent systems ranging from intelligent cities to smart materials. Nikolaus earned his PhD in computer science from EPFL working with Alcherio Martinoli. Before joining EPFL, Nikolaus was a research assistant at the Collective Robotics Group at Caltech. Nikolaus studied electrical engineering at ETH Zürich and TUM, with stints at Lund University and Caltech.

Hosted by James Martin.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)