Approaches to Cluster Formation in Wireless Sensor Networks Jimmy Wilson CS 526
Presentation Outline Assumptions Assumptions Why form clusters? Why form clusters? Clustering objectives Clustering objectives Approaches Approaches Further study Further study
Assumptions Many sensors are deployed randomly. Many sensors are deployed randomly. Sensor mobility is fixed. Sensor mobility is fixed. Homogenous sensors (limited resources). Homogenous sensors (limited resources). Purpose of the sensors is to report temperature, pressure, humidity, light, or chemical activity. Purpose of the sensors is to report temperature, pressure, humidity, light, or chemical activity. Reports will be sent to a base-station. Reports will be sent to a base-station.
Why form clusters? Reduce routing table size. Reduce routing table size. Conserve communication bandwidth. Conserve communication bandwidth. Avoids redundant message exchanges. Avoids redundant message exchanges. Cluster head can aggregate data. Cluster head can aggregate data. Clusters can schedule activities so sensors can sleep. Clusters can schedule activities so sensors can sleep.
Clustering Objectives Energy Efficiency Energy Efficiency Load Balancing Load Balancing Fault-tolerance Fault-tolerance
Clustering Approaches Clustering Approaches Energy Efficient Failure Recovery Balanced Clustering EEHCYesN/AOK LEACHNoYesOK Extended HEEDYesN/AVery good DWEHCYesN/AVery good MOCAYesN/AGood Attribute-based Clustering Yes Very good
MOCA: Multi-hop Overlapping Clustering Algorithm Argues that some degree of cluster overlap can assist with inter-cluster routing, topology discovery and node localization and fail over. Argues that some degree of cluster overlap can assist with inter-cluster routing, topology discovery and node localization and fail over. Algorithm’s goal is to ensure that each node is either a Cluster Head or within k hops (k is preset cluster radius). Algorithm’s goal is to ensure that each node is either a Cluster Head or within k hops (k is preset cluster radius).
Attribute-based Clustering Base-station starts the process by asking nodes to form clusters. Base-station starts the process by asking nodes to form clusters. Cluster Heads are rotated. Cluster Heads are rotated. Fail over is easily detected since Cluster Heads send regular heat beats to the members. Fail over is easily detected since Cluster Heads send regular heat beats to the members. If members do not hear a heart beat, then assume problems with CH and assume the role. If members do not hear a heart beat, then assume problems with CH and assume the role.
Further Study Investigate the impact of cluster formation when the base-station is not at a fixed location. Investigate the impact of cluster formation when the base-station is not at a fixed location. Investigate the impact of allowing cluster overlap with AMPL (Jeffrey Rupp’s paper). Investigate the impact of allowing cluster overlap with AMPL (Jeffrey Rupp’s paper).
Questions?
References Ameer Ahmed Abbasi and Mohamed Younis, A Survey on Clustering Algorithms for Wireless Sensor Networks, June 2007 Ossama Younis, Marwan Krunz, and Srinivasan Ramasubramanian, Node Clustering in Wireless Sensor Networks: Recent Devolopments and Deployment Challenges, June 2006 Jeffrey D. Rupp, AMPL: Active Multi-power Level Cluster Formation, 2006 Taewook Kang, Jangkyu Yun, Hoseung Lee, Icksoo Lee, Hyunsook Kim, Byunghwa Lee, Byeongjik Lee, and Kijun Han, Clustering Method for Energy Efficient Routing in Wireless Sensor Networks, February 2007 Tal Anker, Danny Bickson, Danny Dolev, and Bracha Hod, Efficient Clustering for Improving Network Performance in Wireless Sensor Networks, 2008