2008.04.15 |
| Date | Wed Apr 30 |
| Time | 14:15 — 15:00 |
| Location | DI-eReception |
Title:
Towards Scientific Sensornet Engineering: Lessons from the Hogthrob Project
Abstract:
Sensor networks should allow ecologists to densely sample phenomena that have traditionally been hard to observe in-situ.However, in today's deployments the ratio of computer scientists to ecologists is 3:1 or higher. Even small-scale sensor networks need to be carefully designed, programmed and deployed to meet experimental goals in terms of lifetime or quality of the collected data. Only by confronting the problems that do come up during actual deployments can we devise the tools, methods and abstractions that will make sensor networks an appropriate instrument for scientists. In this talk, I will report on the lessons we learnt in the Hogthrob project, building a sensornet for monitoring group-housed sows. I will, first, describe the sources of complexity in the design of such a system. I will then describehow we coped with this complexity over two iterations. Finally, I will generalize from our experience and discuss design principles for scientific sensornets.
Presenter:
Philippe Bonnet
Associate Professor
Department of Computer Science
University of Copenhagen