]> There is a growing interest in the research community towards a paradigm in which computing is embedded into our daily lives. Everyday objects shall act as computing devices, enriching user experience and enhancing the environments in which we live. Sensor Networks are a step in this direction. Sensor Network research is still in its infancy, with research spanning from hardware to communication to data management and to software engineering principles. To build a strong foundation on the principles underlying sensor networks and enable rapid prototyping,we require development environments to architect these systems and tools to simulate the performance of the algorithms and techniques that we develop.

To this end, MagicWeaver aims to provide a agent-based simulation framework for wireless sensor networks. It provides abstract models for sensor tasking and communication, sensor device contraints, network topologies and the physical environment. It offers the flexibility to incorporate user-defined instantiations of these models into the runtime system. In this paper we discuss the design rationale of MagicWeaver and provide it's current status of the implementation. To experimentally verify the simulation results given by the framework, we compare three data propagation schemes, Flooding, Gossipping and SPIN.

]]> 2001-09-01T00:00:00-05:00 2002-08-01T00:00:00-05:00 2002-08-01T00:00:00-05:00