Geographic routing in wireless sensor networks is based on the prerequisite that every node has information about its current position, for instance via GPS or some localization algorithm. This location information has a certain degree of inaccuracy in real deployments. The majority of geographic routing algorithms, however, has been designed for nodes with exact position information. We show that location errors yield bad performance or even complete failures. Two elaborated geographic routing algorithms for sensor networks, GPSR and BGR, are evaluated with the nodes having inaccurate location information, varying the standard deviation of the position error between zero and the transmission range. Simulation studies show a vast decrease of the packet delivery ratio. To enhance both algorithms, fixes for them are presented to improve the delivery ratio and to save energy in case of location errors. 1.

Wireless sensor networks (WSNs) pose challenges not present in classical distributed systems: resource limitations, high failure rates, and ad hoc deployment. The lossy nature of wireless communication can lead to situations, where nodes lose synchrony and programs reach arbitrary states. Traditional approaches to fault tolerance like replication or global resets are not feasible. In this work, the concept of self-stabilization is applied to WSNs. The majority of self-stabilizing algorithms found in the literature is based on models not suitable for WSNs: shared memory model, central daemon scheduler, unique processor identifiers, and atomicity. This paper proposes problem-independent transformations for algorithms that stabilize under the central daemon scheduler such that they meet the demands of a WSN. The transformed algorithms use randomization and are probabilistically self-stabilizing. This work allows to utilize many known self-stabilizing algorithms in WSNs. The proposed transformations are evaluated using simulations and a real WSN.

Abstract — The paper presents a generic, high-level Java interface for the vertical integration of wireless sensor networks. The intuitive interfaces are implemented in a framework that is easy to use. The classes of the framework can be extended to meet the requirements of a wide range of applications. In particular, the framework supports sending packets to and receiving packets from nodes of the sensor network. Packet types are represented as Java classes generated from meta-data based on XML Schema. This approach fosters short development cycles and provides the productivity needed in vertical integration applications. The ScatterWeb platform is used as a sample platform for sensor networks. 1