Low Duty-Cycled Wireless Sensor Networks: Connectivity and Opportunistic Routing

Abstract

This thesis addresses a number of performance and design issues that arise in a low duty-cycled wireless sensor network.

The advances in sensing technology, miniaturization and wireless communication have led to a large number of emerging applications using networked wireless sensors. One of the most critical design goals is the longevity of the system. A widely accepted and commonly used method of energy conservation is duty cycling -- sensor nodes are periodically put to sleep mode to conserve energy. While effective in prolonging the system lifetime, duty-cycling disrupts communication and sensing capabilities as sensors alternate between sleep and wake modes. This not only affects network coverage and connectivity, but also causes delay in message delivery. A central theme of this thesis is to understand the energy-performance trade-off and design good networking algorithms that work well with low duty-cycled sensors. Our work thus centers on how the performance degradation caused by duty-cycling may be mitigated.

The first method is to add redundancy to the deployment: the more sensors we deploy, the more we can reduce the duty cycle of individual sensors while maintaining the system level performance. In this context we investigate the fundamental relationship between the amount of redundancy required vs. the achievable reduction in duty cycle for a fixed performance criterion. We examine this relationship in the case of asymptotic network connectivity.

A second method is to design good algorithms that effectively deal with temporal loss of connectivity. Within this context, we first develop a routing scheme using an optimal stochastic (also referred to as opportunistic) routing framework, designed to work in the presence of duty-cycling as well as unreliable wireless channels. We then examine how the routing delay of this type of algorithms scales compared to conventional (non-opportunistic) routing algorithms in a limiting regime where the network becomes dense.

Lastly, for any routing algorithm to work properly there needs to be an efficient broadcast mechanism that discovers and disseminates topology information. In this context we develop an analysis-emulation hybrid model that combines analytical models with elements of numerical simulation to obtain the desired modeling accuracy and computational efficiency.