Resource management in energy harvesting wireless sensor networks
Date of Issue2015
School of Electrical and Electronic Engineering
With the emergence of energy harvesting techniques, it is now possible for wireless sensor networks to operate perpetually while supporting certain performance levels. Due to the renewable but non-deterministic nature of the energy harvesting source, the way to manage the harvested energy and provide such perpetual operation becomes a major challenge. Thus, in this thesis, we focus on the energy resource management mechanisms for energy harvesting wireless sensor networks. In order to achieve perpetual operations, an Energy Neutral Management (ENM) mechanism is needed to make sure that the harvested energy will be able to replenish the energy that is being consumed by a sensor. Based on different performance maximization goals, we identify two levels of energy neutral management, namely Node-level ENM and Network-level ENM. For Node-level ENM, we study the ways to efficiently utilize the harvested energy so that a sensor can operate perpetually with desired sensor performance level. We firstly consider the case when the sensor performance level (such as the duty cycle) has a linear relationship with the amount of energy consumed by the sensor. We analytically derive a set of energy allocation principles to maximize the amount of harvested energy that can be utilized by a sensor, in the presence of battery storage inefficiencies. These principles in turn maximize the sensor average duty cycle while maintaining its energy neutral state. Since the energy harvesting information is not always available before sensor deployment, we develop a Prediction-FRee Energy Neutral (P-FREEN) management mechanism to implement the derived energy allocation principles based solely on current observed energy harvesting rate and battery residual energy level, which enables perpetual sensor operation with maximized sensor performance level. We next consider the case when the sensor performance level (such as the communication channel throughput) has a non-linear relationship with the amount of energy consumed by the sensor. An off-line optimal energy allocation mechanism, which maximizes the average channel throughput while maintaining the energy neutral state of the sensor, is developed via convex optimization. Based on this optimal mechanism, we propose an on-line Adaptive Energy Budget Assignment Policy (ABAP) that asymptotically maximizes the average channel throughput by using the historical energy harvesting and channel state information observed by the sensor. We also study a method to reduce the energy loss caused by the battery energy storage inefficiencies. The fraction of the harvested energy that can be utilized by using this method is analytically derived and is integrated into ABAP to provide improved average channel throughput. For Network-level ENM, we study the network layer routing protocols that coordinately control the energy consumption of sensors in the network, (by controlling the routing paths of the data traffic), so that perpetual network operations can be achieved with improved network performance levels. We focus on developing routing protocols based on two widely used data delivery models, namely the Query Driven Model and the Continuous Model. Using the query driven model, only data queried by the user will be sensed and delivered to the destination, which prevents the delivery of non-desired data and is thus highly energy efficient. Hence, we propose a query driven Energy Neutral Directed Diffusion (ENDD) protocol to provide Network-level ENM. ENDD employs the traffic flow admission control mechanism to regulate the traffic load carried by a sensor based on its energy harvesting status, which in turn prevents sensors from shutting down due to excessive usage of energy. In this way, routing path failures can be prevented, which ensures the data delivery consistency and improves the network data throughput. Continuous model is used for applications that require the periodical sensing and delivery of data information. Clustering protocols are suitable for such data delivery model as it can enable in-network data aggregations and thus reduces energy wastage caused by the delivery of the redundant data information. We develop an Energy Neutral Clustering (ENC) protocol to group the network into several clusters with the goal of providing perpetual network operation with consistent data delivery. ENC employs a novel Cluster Head Group (CHG) mechanism that allows a cluster to use multiple cluster heads to share the heavy traffic load. This CHG mechanism can help reduce the frequency of cluster re-formations, which in turn reduces the control message overhead. The optimum number of clusters that maximizes the amount of information gathered from the network is mathematically derived via convex optimization. Based on this optimum number of clusters, an extension to ENC is proposed to group the network into equal sized clusters so that maximized network information gathering can be achieved. The performance of our proposed energy management mechanisms is verified through theoretical analysis and extensive empirical studies. We believe that these mechanisms and their results make important contributions to the study of the energy management mechanisms in energy harvesting wireless sensor networks.
DRNTU::Engineering::Computer science and engineering::Computer systems organization::Computer-communication networks