Cushion: Autonomically Adaptive Data Fusion in Wireless Sensor Networks

Abstract

Existing protocols for in-network fusion of data, in wireless sensor networks, are either single-path (i.e., tree-based) or multi-path. Tree-based fusion protocols have small message overhead but low reliability under node and link failures, while multi-path protocols have good reliability, but potentially higher overhead. This paper presents a suite of protocols, called Cushion, which automatically adapts message overhead in order to maintain a desired level of reliability. As a part of Cushion, we present (1) a simple adaptive aggregation protocol as well as (2) a node-aware adaptive aggregation protocol. Both these protocols use special control messages that enable nodes to decide when to transition between using single path and multi-path fusion approaches. The protocols span a continuous spectrum between these two approaches and can further increase reliability over multi-path approach using redundant transmissions. We present experimental results quantifying the benefits and drawbacks of using a Cushion-based approach versus both tree-based and multi-path approaches.