Sensors are increasingly becoming part of our daily lives: motion detection, lighting control, environmental monitoring, and keeping track of energy consumption all rely on sensors. Combining data from this wide variety of sensors will result in new and innovative applications. However, access to these sensors -or the networks formed by them -is often provided via proprietary protocols and data formats, thereby obstructing the development of applications. To overcome such issues, middleware components have been employed to provide a universal interface to the sensor networks, hiding vendor-specific details from application developers. The scientific literature contains many descriptions of middleware components for sensor networks, with ideas from various fields of research. Recently, much attention in literature is aimed at what we, in this paper, define as 'centralised' middleware components. These components consider sensor networks that have no capacity -in terms of memory, data storage, and cpu power -to run middleware components (partially) on the sensor nodes. Often, viewed from the position of the middleware component, these sensor networks function as simple data providers for applications.In this paper we introduce the term 'centralised' for such middleware components, guided by a literature review of existing middleware components for sensor networks. We describe their general architecture, give a description of a representative set of four centralised middleware components, and discuss advantages and disadvantages of these components.
1Finally, we identify directions of further research that will impact centralised middleware systems in the near future.