The increasing complexity, heterogeneity and reliability requirements of wireless sensor networks is posing major challenges to the capability of developing effective designs. The lack of a system level approach is significantly slowing the adoption of this technology and limiting it to marginal markets.

This Dissertation proposes a new methodology for the system level design of wireless sensor networks. This methodology is based on the Platform Based Design (PBD) methodology that was originally developed for classical embedded systems, and it is here revisited and applied to the wireless sensor networks domain. According to PBD, a design is obtained as a sequence of refining steps that take guide the designer from the initial specification all the way down to a physical implementation. To support this process, a set of intermediate abstraction layers and platforms are identified.

When applying this methodology to wireless sensor networks, three layers of abstraction and relative platforms are identified: a service platform at the application layer, a protocol platform to describe the protocol stacks, and an implementation platform for the hardware nodes. Two different strategies to refine applications into implementations using these three platforms are proposed. This differentiation of the mapping strategies is essential to accommodate different application categories. The proposed methodology is validated using case studies from building monitoring and industrial monitoring applications.