Xu Ning

May 2009
Control and Optimization Approaches for Power Management in Wireless Sensor Networks
Committee Members: Advisor: Christos Cassandras, SE/ECE; Ioannis Paschalidis, SE/ECE; Pirooz Vakili, SE/ME; John Baillieul, SE/ME; Appointed Chair: Sean Andersson, SE/ME

Abstract: AWireless Sensor Network(WSN) is a spatially distributed wireless network consisting of low-cost,battery- powered nodes that have sensing and wireless communication capabilities. Power management is one of the key issues in WSNs because it directly impacts the lifetime of such networks. In this dissertation, we focus on reducing power consumption from two different aspects: transmission control and topology optimization.

Transmission control concerns the transmission and receiving of messages across a link in the network. In this aspect, we propose two approaches to reduce control packet overhead. The first one is Message Batching from the sender’s side, which waits for several messages before sending the batch of messages out in a single transmission. The key problem is how to find out the optimal batching size/time resulting in the best tradeoff in energy vs. performance. We derive analytical results for Markovian systems, as well as using Perturbation Analysis to derive unbiased gradient estimators for on-line optimization. The second approach is Dynamic Sleep Time Control from the receiver’s side. While traditional approaches use fixed sleep time (time between consecutive channel pollings in Low-Power Listening) in sensor nodes for engineering simplicity, it is shown that by varying sleep time dynamically, better performance can be achieved because more statistical information is used. Two ways to control sleep time are proposed, catering to different constraints and objectives. An on-line distribution learning algorithm is also devised for practical implementation.

Topology optimization concerns the problem of how a message is routed from one node to another, and how the network is deployed in order to reduce communication energy consumption. Two approaches are investigated in this dissertation. The first one focuses on routing optimization in a flat-topology network that maximizes network lifetime. We have extended previous results by deriving simpler equivalent formulations. We also incorporate a more realistic battery dynamical model, and formulate/solve a new lifetime maximization problem based on this model. The second one focuses on both deployment and routing on a hierarchical WSN with additional constraints on reliability. While the energy optimization formulation is very difficult to solve, we manage to decompose the problem and solve it iteratively. This decomposition algorithm greatly increases the solving speed and scalability compared to existing commercial solvers. An incremental deployment scheme based on this algorithm is also proposed.