Computer and Information Science November, 2009 3 A Weighted-Density Connected Dominating Set Data Gathering Algorithm for Wireless Sensor Networks Larry King Clemson University, Clemson, SC 29634, USA E-mail: larryfking3@gmail.com Natarajan Meghanathan (Corresponding author) Department of Computer Science, Jackson State University P. O. Box 18839, 1400 John R. Lynch Street, Jackson, MS 39217, USA Tel: 01-601-979-3661 E-mail: natarajan.meghanathan@jsums.edu This research is funded by the U.S. National Science Foundation through grant (CNS-0851646) entitled: “REU Site: Undergraduate Research Program in Wireless Ad hoc Networks and Sensor Networks.” Abstract We propose a weighted-density connected dominating set (wDCDS) based data gathering algorithm for wireless sensor networks. The wDCDS is constructed using the weighted-density of a sensor node, which is defined as the product of the number of neighbors available for the node and the fraction of the initially supplied energy available for the node. A data gathering tree (wDCDS-DG tree) rooted at the wDCDS Leader (the node with the largest available energy) is formed by considering only the nodes in the wDCDS as the intermediate nodes of the tree. The leader node forwards the aggregated data packet to the sink. The wDCDS and wDCDS-DG tree are dynamically reconstructed for each round of data gathering. Simulation studies reveal that the wDCDS-DG tree yields a significantly larger network lifetime, lower delay and lower energy consumption per round compared to the density-only CDS and energy-only CDS based data gathering trees. Keywords: Connected Dominating Set, Density, Energy, Data Gathering Tree 1. Introduction A wireless sensor network is a distributed system of smart sensor nodes that collect data about the ambient environment and propagate the data back to one or more control centers called ‘sinks’, which access the data. A sensor node typically has limited battery charge, computing capability and memory capacity. The transmission range of a sensor node is the distance within which the signals emanating from the node can be received with appreciable signal strength. Wireless sensor networks have limited bandwidth as the sensor nodes within the transmission range of each other share the communication medium. As the sink is normally fixed and located far away from the sensor network field, direct transfer of the collected data to the sink is not a viable solution from both energy as well as bandwidth point of view. This motivates the need for data gathering algorithms that can be effectively and efficiently run at the sensor nodes to combine the data and send only the aggregated data (that is a representative of the data collected from all the sensor nodes) to the sink. Throughout this paper, the terms ‘data aggregation’, ‘data fusion’ and ‘data gathering’ are used interchangeably. They mean the same. Data gathering algorithms typically proceed in rounds, wherein during each round, data from all the sensor nodes are collected and aggregated, and then forwarded to the sink. The communication structures normally used for such data aggregation are clusters (Heinzelman, Chandrakasan & Balakrishnan, 2004), grid (Luo, Ye, Cheng, Lu & Zhang, 2005), chain (Lindsey, Raghavendra & Sivalingam, 2002), connected dominating set (CDS) (Meghanathan, 2009) and trees (Lindsey, Raghavendra & Sivalingam, 2001). Meghanathan (2009) proposed an energy-based algorithm to construct the CDS (called the ECDS) of the underlying sensor network for every round of communication and also to construct a data gathering tree (ECDS-DG tree) based on the ECDS. The ECDS strategy prefers to include nodes that have a relatively high energy as intermediate nodes of the data gathering tree (i.e., used for data collection, aggregation and forwarding);