The case for interpreted languages in wireless sensor networks

Abstract

As sensor networks gain popularity and technology scaling allows further processing in each network node, the programming of these distributed computational structures becomes a serious bottleneck. Interpreted languages adoption may allow a smaller programming effort, and since they show a denser code representation than their directly executed counterpart, interpreted code exhibits smaller power dissipation during over-the-air reprogramming. As technology scales, the processing energy cost tends to reduce more than communication energy, which is bounded by the required irradiated radio power. By allowing the execution of more complex software WSN can be used for more refined applications, like image processing, compression and recognition. Also, interpretation can allow the use of object oriented technology software, allowing high productivity gains. However, the interpretation overhead cost and the extra memory required in Java, for example, argue against interpreted languages adoption in WSN. In this paper we show the design space for interpreted languages, and demonstrate that there is a large application domain where interpretation benefits can be used together with energy efficiency