Design, Analysis and Empirical Evaluation of Power Management in Multi-Hop Wireless Networks

Energy is a primary resource in multi-hop wireless networks with small devices powered by battery. Minimizing energy consumption without affecting communication activities is crucial to prolong the lifetime and improve robustness of the network. One effective way to reduce the energy consumption of wireless devices, termed power management, is to put the entire or part of the system in the low-power state. In this thesis we develop a comprehensive set of solutions to power management in multi-hop wireless networks from protocol design, performance analysis to empirical evaluation. The main objective is to achieve energy conservation with moderate and controllable performance degradation. We first propose an on-demand power management framework that utilizes the short-term dependency in packet arrivals and the intrinsic correlation between on-demand routing protocols and data delivery to achieve energy saving, while not introducing significant delivery latency. Both simulation studies and empirical evaluation demonstrate significant energy savings of on-demand power management. To counter the problems of a centralized time server and poor adaptability to network dynamics in synchronized wakeup schedules (e.g., IEEE 802.11 PSM), we then design an optimal asynchronous wakeup schedule based on the theory of block design. Two power management schemes, i.e., slot-based and on-demand power management are investigated and evaluated in conjunction with the proposed asynchronous wakeup protocol. In the third part of the thesis, we develop queuing models to characterize energy consumption, delay and throughput as a function of wakeup schedule and power management strategies, and analyze the energy-performance trade-off in power managed networks. We also implement several modules and APIs in TinyOS 1.0 for time-synchronization, profiling the energy consumption and facilitating power management of small wireless devices. The effectiveness of the proposed power management protocols are corroborated by the measurement-based study using UCB motes