Service Clusters

Type: Semesterarbeit
Students: Roman Amstutz
Advisor: Clemens Lombriser
Project: e-SENSE

Personal Area Sensor Networks are  networks consisting of several nodes on one person. If there are several persons in a certain area, the nodes of the different Personal Area Sensor Networks can hear each other and build up communication. Since persons move around, the whole network topology can quickly change. One approach to adapt to this changing environment is to build subnetworks. Each of these subnetworks contain nodes which stay together for a longer period of time. The process of grouping nodes to a subnetwork is called clustering. In our approach a cluster is always managed by a single node, the clusterhead.

Our thesis presents a clustering protocol that deals with the specific mobility requirements of a body-worn system. Additionally a simple service discovery protocol generates a service directory for the cluster on one node. The protocols are implemented as applications in TinyOS, which is an open source operating system for wireless embedded sensor networks.

Different simulation models are presented and used to evaluate the protocol against mobility requirements. The behavior of the protocol in selected situations is depicted to demonstrate the operation of the protocol. The protocols were evaluated by using a simulator for TinyOS called TOSSIM. TOSSIM is programmed in tython, a scripting language quite similar to python. It is able to simulate several sensor nodes in a network and allows to select different connectivity models for the radio transmission.

We managed to improve the original protocol by decreasing the number of messages generated for clustering. Additionally, we give recommendations on choosing important protocol parameters.