66
WIRELESS SENSOR NETWORK COVERAGE:
DEMONSTARTING POWER SAVINGS AND LIMITATIONS WITH MINIMIZED COVERAGE
BY
SIDNEY I. RUBEY
A thesis submitted to the Graduate faculty of the
University of Colorado at Colorado Springs
in partial fulfillment of the
requirements for the degree of
Master of Science
Department of Computer Science
2011
This thesis for Master of Science degree by
Sidney I. Rubey
has been approved for the Department of Computer Science
by
------------------------------------------------------------------------
Dr. C. Edward Chow, Chair
------------------------------------------------------------------------
Dr. Terrance E. Boult
------------------------------------------------------------------------
Dr. Xiaobo Zhou
-----------------------------------------
Date
Abstract
This paper explores the competing issues of coverage efficiency and power available in wireless sensor networks. Specifically, a shortest distance routing protocol is implemented, and total network system lifetimes are determined using a variety of small percentages of the available system nodes. Using a network simulator developed in Java, wireless sensor nodes are simulated, and power consumption algorithms are included in each node that take into consideration all aspects of power consumption in the operation of the node. Simulating different coverage schemes on the same network system, same initial power sources, and routing protocol, an increase of overall system lifetime of 900% is demonstrated, but also that the network lifetime increase does not increase linearly as the percentage of nodes used in the system is decreased.
Acknowledgement
I would like to thank Professor C. Edward Chow for his guidance and patience during my thesis work. I would also like to thank my family for their support, patience, and understanding during this labor when I had to be “working on my thesis”.
Abstract iii
Acknowledgement iv
1 INTRODUCTION 1
1.1Wireless Sensor’s Structure and Operation 1
1.2 Research Goal: Demonstrating Advantages of Reducing Coverage 3
1.3 Paper Organization 4
2 RELATED WORK 6
2.1 Power 6
2.2 Recent Developments with Power 8
2.3 Coverage 9
2.4 Wireless Sensor Network Simulators 13
3 PRELIMINARIES 16
3.1 Battery, Power Consumption & Management 16
3.2 Routing Protocol 18
3.3 Placement of Sensors 20
4 IMPLEMENTATION OF SIMULATION 21
4.1 Assumptions 21
4.2 Java Network Simulator 21
4.3 Power Consumption 29
4.3.1 Microcontroller 30
4.3.2 Sensing Unit 30
4.3.3 Radio 30
4.4 Coverage 30
5 ENHANCEMENT 1 32
6 ENHANCEMENT 2 34
7 COMPLEXITY ANALYSIS OF ALGORITHM 35
8 EXPERIMENTAL RESULTS 37
8.1 Initial Simulation Configuration 37
8.2 Enhancement 1 41
8.3 Enhancement 2 43
9 LESSONS LEARNED 45
10 CONCLUSIONS AND FUTURE RESEARCH 47
Appendix A 54
Appendix B 58
Appendix C 59
Figures
Figure 1: Berkeley’s SmartDust, Crossbow Technologies MICA2…………………2
Figure 2: Converting RF energy to DC power………………..................................9
Figure 3: Algorithm of MUEL Simulator................................................................21
Figure 4: Display of output window with nodes near each other............................30
Figure 5: Graph of Network Life vs. Number of Subgroups...................................40
Tables
Table 1: Current used in Crossbow MICA2 Mote...........................................................25
Table 2: Network lifetimes, first configuration................................................................35
Table 3: Network lifetimes, second configuration...........................................................36
Table 4: Network lifetimes with Enhancement #1...........................................................38
Table 5: Network lifetimes with Enhancement #2...........................................................39
66
Chapter 1
1 INTRODUCTION
Wireless sensors, and the arrangement of these small, electronic devices into radio networks, have introduced the capability of remotely monitoring a physical environment for a wide variety of parameters. In effect, there is additionally the ability to interface the physical world, monitoring whatever parameter is of interest, with the internet. A scientist may, for example, be able to view conditions in a rain forest on another continent, from his office in Colorado. [LE10]
1.1Wireless Sensor’s Structure and Operation
Wireless sensors are devices that range in size from a piece of glitter to a deck of cards. They are functionally composed of:
· a sensing unit that is designed and programmed to sense whatever characteristic is of interest; some common examples of properties that are monitored are light, temperature, humidity, pressure, etc.
· a converter that transforms the sensed signal from an analog to a digital signal;
· a microprocessor controlling component that includes an operating system for the unit, processor and memory;
· a radio component that includes both a receiver and a transmitter.
Powering these components is typically one or two small batteries. There are also wireless sensors utilized in applications that use a constant, wired power source and do not use batteries as a power source. This type of wireless sensor is not considered in this paper.
Figure 1:Berkeley’s SmartDust [PI01] Crossbow’s Mica mote (appx. 60 mm across) [XB10]
In an external environment where the power source is batteries, which this paper will concentrate on, wireless sensors are placed in an area of interest that is to be monitored, either in a random or known fashion. The sensors self-organize themselves in a radio network using a routing algorithm, monitor the area for whatever parameter it was designed to monitor, and transmit the data to a central node, sometimes called a base station, or sink node, that collects the data from all of the sensors. This node may be the same as the other nodes, or because of its increased requirements, may be a more sophisticated node with increased power. The unique advantage of wireless sensors is that they may be deployed in an environment for extended periods of time, continuously monitoring the environment, without the need for human interaction or operation. This, however, establishes the power source as the limiting component of the sensor.
1.2 Research Goal: Demonstrating Advantages of Reducing Coverage
Thus introduces the tradeoff between power and coverage in wireless sensor networks. A system of wireless sensor nodes organized into a network may turn on every node for maximum coverage, sometimes introducing redundant coverage over an area, and the lifetime of the system will be minimized. Maximum, redundant coverage may be appropriate and desired in some applications, such as using tiny sensor nodes to monitor a critical health situation in humans, and also in situations where replacement of the batteries is possible when power is completely depleted. Maximum, redundant coverage may also be desirable in applications using rugged, environment sensor nodes for monitoring a strategic military scenario for a known duration of time. In these examples, the lifetime of the system is not maximized, and this is an entirely acceptable consequence. At the other end of the spectrum, however, there are applications where maximum, redundant coverage is not the paramount consideration. A reduced coverage scheme may be wholly acceptable, as a tradeoff for vastly extending the lifetime of the system. For example, any phenomenon that is not too divergent throughout an area; events, phenomenon that are somewhat continuous and one event at one point is not drastically different than the same monitored event at an adjacent point. Put another way, in applications where the tradeoff of coverage to gain extended system lifetime does not have drastic negative consequences. Some examples of these types of phenomenon are temperature, light, sound, atmospheric conditions.
The advantages of minimizing coverage, under appropriate circumstances, are the focus of this paper. For an area that is being monitored by a wireless sensor network, with nodes having a finite, known sensing range, and turning on one node in an area that produces no overlap in coverage, establishes one “spot” of coverage in the area. Hence, the term “spotty” coverage is sometimes used to refer to fewer sensor nodes being turned on with no overlap.
For example, a wireless sensor network may have a calculated lifetime of 100 hours using 100% of the sensors. If 50% of the sensors are used at one time, while the remaining 50% are reduced to a low-power sleep state, it would be expected that the overall lifetime of the system would be approximately doubled to 200 hours, and this is what the simulation demonstrates.
1.3 Paper Organization
The advantages of minimizing coverage in varying degrees, under appropriate circumstances, are the focus of this paper. The remainder of this paper will cover the following areas: First, a review of research that has already been done that relates to this topic will be covered. This will cover a review of papers that deal with minimizing power consumption – or extending network life - in wireless sensor networks; then, papers that deal with the sensing coverage of wireless sensor networks; and finally wireless sensor network simulators. Specific attention is paid to research that dealt with minimizing the coverage of wireless sensor networks to extend network lifetime. The next section will introduce the basics of electricity and how power consumption is measured in the wireless sensor nodes. Following this, will be a discussion of the routing protocol that has been implemented in the network simulator, which will lead to a discussion of the network simulator that has been implemented using the programming language Java. Finally, the experimental results obtained from simulating power consumption in wireless sensor networks under different conditions will be presented. This will be followed with lessons learned and conclusions to be drawn from the research.
Chapter 2
2 RELATED WORK
2.1 Power
The first topic in the “Related Work” section has to do with the power usage of the wireless sensor nodes. This has often been considered a function strictly of transmission – what routing protocol or algorithm is being used for sending data, messages, network traffic, etc. As I demonstrate in this paper, and is also documented in some of the following papers, energy is also consumed by the other components of the node, and this can be significant. Another distinction to be made on this topic is that some papers are focused on minimizing power usage, by whatever means – and thereby extending network life, and some papers focus on maximizing network lifetime.
The work by Schurgers, et al, “Energy Efficient Routing in Wireless Sensor Networks”, discusses two approaches to making the most efficient use of limited energy in sensors, and thereby extending the life of the network system. The first approach is to use a concept they termed Data Combining Entities, or DCE’s. This concept is similar to clustering, but it does not specifically designate a cluster head; instead it picks a node that has other streams of network traffic flowing through it as the DCE. In this manner, nodes that are in close proximity to the DCE relay their packets to the DCE which can compress and then forward these nodes’ packet for them. Their simulations demonstrate that energy consumption using this method can be reduced by a factor of 2 to 3. [SC01]
The second technique demonstrated by Schurgers to reduce energy consumption in a wireless sensor network is the spreading of network traffic over the entire network. This is opposed to network traffic passing through a few critical nodes, which is typical of how network traffic with randomly placed nodes normally evolves. The energy in these few critical nodes understandably is depleted quickly as it passes other nodes’ message traffic.
“The idea is to divert new streams away from nodes that are currently part of the path of other streams. A node that receives packets tells all its neighbors, except to the one from where the stream originates, that its height has increased.” [SC01]
The “height” is another term for number of hops to the base node. In this manner, it is telling its neighbors that it is not an efficient route to the base. By attempting to spread the network traffic over more of the network in this manner, simulations demonstrate the network remains intact 90% longer than a stochastic routing protocol. [SC01]
Slijepcevic, et al, in the paper “Power Efficient Organization of Wireless Sensor Networks”, focuses on reducing the overall power in the network system by grouping the sensor nodes into mutually exclusive sets. In this manner, the coverage throughout the surveillance area is maintained by the sensor nodes, and turning on only one of sensor nodes of the set at a time. This technique assumes that sensor nodes are placed stochastically. [SL01]
There are many more research topics that demonstrate energy, or power, saving, efficiency, optimization in wireless sensor networks. The subjects range from, but are not limited to, energy conservation with regard to routing protocols, operating systems, software systems, grouping or clustering nodes for power savings, placement of nodes, energy efficient electronic and radio transmission devices, and communication techniques. In short, every component and operation of the individual nodes and the network as a whole has been researched for optimum energy usage. [CH01]
“Maximizing System Lifetime in Wireless Sensor Networks” by Dong, is one of the first papers to differentiate between the “time” and “transmission” approaches to overall lifetime of a wireless sensor node network. Dong refers to these as the “time based model” and the “packet based model”. This research considers many different time based models, and also packet based models; the purpose of the paper is not to put forth scheme that is supposed to be the best. Rather, considering different combinations of models, it does an analysis of the models and determines the complexity of each of the models. In the time based model, it is shown that the problem of extending network lifetime while maintaining connectivity the complexity is NP-hard. In the packet based model, it is demonstrated that all models are NP-hard, with the exception of “cases where each node has a fixed transmission power, many-to-one unicast life time, one-to-many unicast lifetime, and one-to-one unicast lifetime are polynomially solvable; also, many-to-many unicast lifetime is also polynomially solvable in the single commodity model”. [DO05]
The next paper considered is “Minimum Power Configuration for Wireless Communication in Sensor Networks”, by Xing et al. This paper approaches the power / network lifetime problem as actually two problems: one, minimizing the number of active nodes in the network to only those required for coverage and connectivity; two, adjusting the configuration of the power consumed for transmissions by the nodes. This paper demonstrates that the optimum power configuration depends on the data rates of the network. Furthermore, it is demonstrated that problem of solving the minimum power configuration is an NP-hard problem. Also, there are two protocols presented: the minimum power configuration protocol (MCPC), and minimum active subnet protocol (MACP). Unique to MPCP, the protocol will dynamically reconfigure the network power usage configuration based on the current data rates. [XI05]
Chao-Lieh Chen et al authored the paper “Energy-proportional Routing for Lifetime extension of Clustering-Based Wireless Sensor Networks”. In this research paper, an algorithm is presented that will determine the energy usage for nodes in an upcoming round of data collection and transmission; it then determines if a cluster-head or a node should be used for forwarding tasks or transmit data to intermediate hops. In this manner, the algorithm helps to use energy evenly throughout the network. Testing and mathematical proofs validate that network lifetime can be extended by dissipating energy evenly throughout the network. [CH07] In addition to Schurgers [SC01], this is another paper that promotes energy being used evenly throughout the entire network will extend overall network life.