The development of increasingly complex algorithms for sensor networks has made it difficult for researchers to implement their design on typical sensor network hardware with limited computing resources. The demands on hardware can also mean that small microcontrollers are not the ideal platform for testing computationally and?or memory intensive algorithms. Researchers would also like access to high level programming languages and a wider range of open source libraries. To address this problem we have designed and implemented an architecture, Gumsense which combines a low power micro-controller (8MHz MSP430) with a powerful processor (100-600MHz ARM) on a Gumstix board running Linux. This Open Embedded OS supports a wide variety of programming languages, package management and development tools. A similar hybrid approach was also used in the LEAP platform. The microcontroller wakes up frequently to manage tasks such as activating sensors and gathering data. The intended use-case is to power-up the ARM board and storage only during the brief periods it is needed, for example performing computation or communication.
pubs.doc.ic.ac.uk: built & maintained by Ashok Argent-Katwala.