Energy efficient sampling, source coding, and data routing in wireless sensor networks

One important problem in wireless sensor networks is how to efficiently utilize the limited network resources to observe and estimate physical phenomena. In this dissertation, we take a divided approach to this problem and separately devise efficient algorithms for field sampling, source coding, and data routing.; Before we start, various types of distortion during the process of sensing, quantization, communication, and reconstruction are examined. It is observed that the bounds on these errors are fundamentally tied to the scarce network resources, e.g. node density, sensor energy, and communication capacity.; Nodes with limited and controlled mobility have been proposed recently for use in wireless sensor networks. The problem of efficiently relocating sensors to sample a distributed field is investigated. We propose an adaptive algorithm based on the Bayesian framework. This scheme maintains an estimate of how well the current reconstructed field approximates the true field based on all collected samples, while iteratively sampling the field by picking the most desirable set of sampling sites from a candidate pool. With minor modifications, this method can also be used in a distributed implementation where static sensors are woken up from sleep to collect measurements.; Due to the high data correlation in a sensor network, source coding should be used to remove redundancy among data streams from different sensors even before they are transmitted to the fusion center to reduce communication cost. We give a brief overview of distributed source coding, where sensors independently conduct data compression without interacting with one another. Then, our attention is shifted to source coding with explicit side information. A two-stage DPCM (differential pulse coded modulation) coding scheme is proposed. It can continuously monitor the additional coding gain provided by correlated side information from other sensors, and hence can be used in joint data aggregation/routing optimization.; The last topic we take up is the data-centric routing. We proposed a data aggregation model for source coding with explicit side information. In this model, data transmissions are decomposed into individual flows originating at different sensors, and a data rate function is defined for each flow. The full optimization problem is formulated and discovered to be NP hard, which indicates that efficient algorithms for finding the exact solution are unlikely to exist. We turn to heuristics subsequently. Several routing schemes are examined. Among them, the BAS (balanced aggregation scheme) and DSIT (designated side information transmission) hold the highest promise as they yield good performance when data correlation is high and converges to SPT (shortest path tree) when coding gain diminishes.