Energy Efficient Wireless Sensor Network Architecture for Flexible Prototyping Juan P. Leal Licudis, Juan C. Abdala, Guillermo G. Riva*, Jorge M. Finochietto Universidad Nacional de Cordoba, Digital Communication Lab, Cordoba, *Universidad Tecnologica Nacional, Department of Electrical Engineering, Cordoba leal.licudis@gmail.com,jcabdala@ieee.org,griva@scdt.frc.utn.edu.ar, jfinochietto@efn.uncor.edu Abstract. The development of sofisticated and energy-efficient mech- anisms and the consequent increase of application complexity in wire- less sensor networks (WSNs) requieres an open and flexible access to the communication stack. Nowadays, researchers in WSNs are focus on not only in the development of mechanisms in application level but also in the interaction with the communication stack in order to improve the performance. Because communication is the most consuming energy component of WSNs the main challenge is to reduce the communication cost by means of more efficient processing distributed in networks. In this work we compare and analize different embedded operating sys- tems (OS) and communication stacks available for WSNs, and propose a real implementation of a probabilistic routing mechanism using a light communication protocol over a versatile operating system. This imple- mentation enables the flexible prototyping of efficient and novel process- ing algorithms in sensor networks. Keywords: Wireless sensor networks, embedded operating system, com- munication stack, probabilistic routing 1 Introduction A WSN consists of spatially distributed autonomous battery-powered sensor nodes which can sense physical parameters of their local environment (e.g. tem- perature, humidity, light intensity, etc) and send this information to a sink or coordinator node via multi-hop wireless communication. Sensor nodes have lim- ited computation, storage and communication capabilities. The energy cost of transmitting 1 bit is approximately equivalent to compute 2000 code instruc- tions. Since communication cost is much higher than computation one in terms of energy, it is preferred to implement in-networks processing tasks to reduce the message exchange [1]. This tasks must be computationally simple because the constrained resources of this technology. Most works in WSNs focus on the development of efficient mechanisms and algorithms validated only by simulation (eg. by using Omnet++, NS2, etc) and