Decision-Making in Wireless Sensor Networks

We consider challenges associated with application domains in which a large number of distributed, networked sensors must perform a sensing task repeatedly over time. We address issues such as resource con- straints, utility associated with a sensing task, and achieving global objec- tives with only local information. We present a model for such applications, in which we define appropriate global objectives based on utility functions and specify a cost model for energy consumption. In the full version of this paper, we present algorithms and experimental results for this problem domain (2). I. INTRODUCTION In this paper, we argue that for many interesting applica- tions of wireless sensor networks, a best-effort service model, in which nodes are expected to perform sensing operations and route data as best they can, may be too stringent. We adopt an economic decision model in which an activity is performed if its associated benefit outweighs its opportunity cost. A significant challenge here is the distributed nature of nodes in our networks, which implies that they do not have global information, making it unrealistic to expect nodes to accurately assess either the op- portunity costs, or the relative benefits of a particular decision. Therefore, nodes in our model make heuristic assessments based on available local information in an attempt to optimize global objectives (1). For the objectives we seek to address, computation in large-scale sensor networks will require scalable coordination amongst sensors to accomplish the desired tasks (3). We con- sider global objective functions motivated by specific sensor network applications which are driven by utility functions, first studied in a networking context by Shenker (5). Developing so- lutions which achieve these objectives are constrained in two primary ways: by the locality imposed by the distributed nature of the model, and by a resource constraint, namely the finite en- ergy supply at sensor nodes. Our work develops a general model in which to study such problems and presents algorithmic results and experimental work in progress for a class of these problems. While the objective functions and algorithms we propose are novel, they connect to a substantial body of work on ad-hoc rout- ing protocols, fault tolerance, and energy conservation in sensor networks, which we survey in the full version of the paper. One work which considers several of the issues we consider here, in- cluding sensor fusion, or aggregating sensory information from multiple sources; load-balancing; and power conservation is the This work was partially supported by NSF research grant ANIR-9986397. Low-Energy Adaptive Clustering Hierarchy (LEACH) protocol (4). The LEACH protocol (4) uses sensor fusion to compress datasets within the network, reducing the energy dissipated dur- ing the resulting transmission. One application-specific example they describe is beamforming algorithms, which combine a set of acoustic signals into a single signal without loss of relevant information. Our work applies the same general principle in ad- vocating application-specific data aggregation as a technique for conserving energy.