SensorNet Paper Reviews by wj69
From Ece
- Return to the paper reading schedule or the Sensor Networks homepage.
Literature Reviews
Connecting the Physical World with Pervasive Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The article presents theories of ideal sensor networks by addressing challenges and surveying current advancements in various fields to achieve ubiquity and invisibility for sensor networks. In order for sensor networks to achieve an “all-knowing” state, the authors address the challenges of technology such as size, power efficiency, wireless communication, and complete autonomy. The article mentions alternative solutions of extracting power from mechanical sources such as movement. In addition to these technological challenges, the authors also address the behavioral issues of sensor networks, and their impact on the observed environment. Behavioral in the sense that sensor networks should have minimal impact on the environment while collecting data. For all the issues addressed, the authors did not address the issue of securing information. The authors present the case of numerous devices monitoring “civil infrastructure.” By detecting weaknesses and transmitting the information, a possibility exists this information can be used for harm. Overall, the authors call for a need for interdisciplinary study as presented by the issues throughout the article, but even with the issues presented, the issue of security should be presented along with the other forefront issues so that it is tightly integrated into the architecture of sensor networks instead of pieced together as an afterthought.
- D. Estrin, D. Culler, K. Pister, and G. Sukhatme, "Connecting the Physical World with Pervasive Networks," IEEE Pervasive Computing, pp. 59-69, January-March 2002.
Sensor Networks for Emergency Response: Challenges and Opportunities
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
In this paper, the authors address general challenges in sensor network design for emergency response by presenting an infrastructure designed by them known as CodeBlue. The problems of power, security, routing, and device size are all addressed in addition to some application specific issues for emergency response. The authors present the CodeBlue infrastructure as a secure, low-power, and robust architecture targeted towards emergency response. From the paper’s explanation, the CodeBlue architecture has viable potential for wide scale deployment based upon the issues addressed. The architecture emphasizes decentralization so as to avoid a single-point of failure. Device size is also addressed since most of the devices will be used in monitoring vital signs, so a large device is not preferred for the wearer. The issue of security has a starting foundation with the authors noting that full scale PKI found in larger systems is not feasible due to limitations of memory and the concept of decentralization of the devices. With the issues presented by the authors, the CodeBlue architecture seems like a promising suite, but the paper could present a little more specific information about the architecture such as the ad hoc routing protocol and node discovery methods used.
- K. Lorincz, D. Malan, T.R.F. Fulford-Jones, A. Nawoj, A. Clavel, V. Shnayder, G. Mainland, S. Moulton, and M. Welsh, "Sensor Networks for Emergency Response: Challenges and Opportunities," In IEEE Pervasive Computing, Special Issue on Pervasive Computing for First Response, Oct-Dec 2004.
Habitat Monitoring with Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper presents the problem of monitoring life patterns of living organisms in the wild. Life patterns include population shifts, habitat needs, etc. and its impact on the ecosystem and vice versa. The paper touches on typical issues of sensor networks such as routing, power, size, communication and offers some solutions currently being researched for these areas. Some of the strengths presented by the authors include utilizing tree-based routing and TDMA schemes for message traffic and communication respectively, but they neglect to mention the direct impact on power. The authors describe the effects, but no hard results are presented. On the other hand, the authors mention the need for durability and need some type of autonomy to recognize when a node is malfunctioning. Their proposed solutions of recognizing packet loss and clock skew seems like a viable solution, but the authors do not explain much more than just mentioning the two methods. For a habitat monitoring system, the ideas presented in this paper place it on the right track, but failure diagnosis and tolerance of the system needs to be presented a little more.
- R. Szewczyk, E. Osterweil, J. Polastre, M. Hamilton, A. Mainwaring, and D. Estrin, "Application Driven Systems Research: Habitat Monitoring with Sensor Networks", Communications of the ACM, Special Issue on Sensor Networks, June 2004.
Towards a Sensor Network Architecture: Lowering the Waistline
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The problem presented by the paper is the wide-ranging protocols, suites, algorithms, etc. for sensor networks. The multitude of wide-ranging protocols, suites, etc. show far areas of research, but also increase the time to achieve the primary goals of sensor networks as well as incompatibility with each other. The paper argues a sensor network architecture must be defined in order reach the goal of viable sensor networks in a reasonable timeframe. This architecture would be similar in what the Internet architecture achieved but also different in implementation due to its nature of application. The paper takes the first step in defining this architecture by proposing the Sensor-net Protocol (SP). The purpose behind SP is to promote advancement while defining a baseline operation. According to the paper, SP is a flexible layer that provides certain services to the other layers. By introducing this concept of a flexible layer in an architecture, researchers can develop sensor networks for wide-ranging applications without feeling heavily restricted and promoting collaboration among other sensor networks.
- D. Culler, P. Dutta, C.T. Ee, R. Fonseca, J. Hui, P. Levis, J. Polastre, S. Shenker, I. Stoica, G. Tolle, and J. Zhao, "Towards a Sensor Network Architecture: Lowering the waistline," Proceedings of Hot Topics in Operating Systems (HotOS '05), 2005.
MANTIS OS: Am Embedded Multithreaded Operating System for Wireless Micro Sensor Platforms
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper presents another alternative for embedded operating systems with the design constraints wireless micro sensors in mind. Current embedded OS systems either use too much power or cannot be dynamically reconfigured without some physical access. Other embedded OSes such as the TinyOS are event based.
MANTIS OS (MOS) is an alternative with the basis of a multi-threaded time-sliced embedded OS with the design limitations of power and memory in mind. The size of MOS fits within 500 bytes of RAM compared to that of μCOS and AVRX, which both have a size just short of 4KB.
MOS is also based on the C language to provide cross-platform support. Instead of needing the hardware to test with, applications for MOS-based sensors can be developed on x86 platforms and ported over to hardware sensors with little or no changes in the code. Another key feature provided by MOS is the deployment of a virtual sensor network as a testbed with the added feature that the virtual sensor network can communicate with the physical sensor network with some limitations.
In order to achieve the stated power requirements, optimizations in MOS are tweaked in the various layers (i.e. scheduler and comm layers). The authors have modeled the sleep functions with that of UNIX architectures to provide a familiar environment when developing applications for MOS. The comm layer is interrupt-driven so as to conserve energy.
Dynamic reprogramming is another stated key feature from the paper. The goal is to completely reprogram the node while deployed, which means in most situations physical access is almost impossible. The paper states that this is a work in progress, but progress is being made either through virtual machines or binary updates.
The paper’s presentation of MOS is very convincing with solid explanations for most of the features stated. Optimizations in the network stack, scheduler, and comm layer provide power efficiency, and the MOS abstractions provide MOS with a fairly small memory requirement. Based from the paper, MOS seems like a solid player in the embedded OS arena.
- S. Bhatti, J. Carlson, H. Dai, J. Deng, J. Rose, A. Sheth, B. Shucker, C. Gruenwald, A. Torgerson, R. Han, "MANTIS OS: An Embedded Multithreaded Operating System for Wireless Micro Sensor Platforms," ACM/Kluwer Mobile Networks & Applications (MONET), Special Issue on Wireless Sensor Networks, vol. 10, no. 4, August 2005.
The Emergence of Networking Abstractions and Techniques in TinyOS
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper is a general survey of various research projects using TinyOS in order to abstract general categorizations. The abstraction categories include general, specialized, in-flux, and absence. The general category includes widely adopted implementations with TinyOS as providing the foundation for mechanism and policy. The specialized category is similar to the general category except policy is specified by the application instead of TinyOS. In-flux includes applications with the mechanisms and policies in the application itself instead of managed by TinyOS. Absence category is more published theory than an actual implementation of the concept. The paper also identifies four common techniques that TinyOS projects tend to converge upon: cross-layer support, static resource allocation, snooping, and scheduled communications. These techniques provide the channels needed to develop applications for resource limited sensor nodes by opening research for two important areas in sensor network research: power and memory. The ideas presented in this paper provide a high level outlook in sensor network research. Even though the paper focuses primarily on TinyOS, the network abstractions and techniques can be abstracted to other embedded OSes like MANTIS OS, which becomes very useful as sensor networks converge upon a common framework for various layers.
- P. Levis, S. Madden, D. Gay, J. Polastre, R. Szewczyk, A. Woo, E. Brewer and D. Culler, "The Emergence of Networking Abstractions and Techniques in TinyOS," USENIX/ACM Symposium on Networked Systems Design and Implementation (NSDI), 2004.
A Dynamic Operating System for Sensor Nodes
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
SOS is another alternative in the wireless sensor embedded OS arena along with TinyOS, Mate, and MANTIS OS. The purpose of SOS is to provide a dynamic operating system that utilizes modules that can be simply plugged into the OS and be ready to go without the need to perform heavy configuration outside of the core development of the module. With the introduction of dynamic operation, the SOS’s goals include being on a level playing field if not better in terms of energy.
Another solution presented by the authors is to perform code updates with minimal energy cost. Because of the architecture of SOS, code updates are significantly smaller compared to TinyOS and Mate, which reduces the amount of energy needed to perform the code update. As noted in the paper, the update cost has minimal effect on overall total energy usage in the long run.
Overall, the SOS seems like a good solution since the results presented show that it’s on a level playing field with TinyOS in terms of energy with the added feature of dynamic resource allocation in a constrained environment. In terms of performance on the dynamic operating system, a comparison to MANTIS OS would be nice because the paper briefly mentions about not having the context switching overhead.
- C.-C. Han, R.K. Rengaswamy, R. Shea, E. Kohler and M. Srivastava, "A Dynamic Operating System for Sensor Networks," International Conference on Mobile Systems, Applications, and Services (MobiSys), 2005.
Just-In-Time Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper presents the need to develop cost-effective wireless sensors to be used for rapid deployment. The solutions provided by the authors have an ideal tone of how wireless sensors need to work without providing concrete evidence of research projects pursuing various endeavors the paper calls for. Some of the ideas include deploying these nodes with a cannon that will program, target, and activate nodes upon deployment. Another idea is to reduce redundancy when deploying, which makes some sense in the cost arena, but if the goal is to deploy cost-effective solutions, redundancy seems to be a necessity especially when deploying in remote regions. The tone of the paper idealizes a sensor network that can perform arbitrary computations according to the application and route the data where needed without really addressing cost factors like power, memory, physical size, etc. that current research is trying to address. The paper is a start for looking at areas to focus on when researching wireless sensors, but needs additional information to provide some concreteness to its idealized tone.
- G. Yee, B. Shucker, J. Dunn, A Sheth, R Han, "Just-In-Time Sensor Networks," IEEE Workshop on Embedded Networked Sensors (EmNets) 2006, pp. 6-10.
Medium Access Control With Coordinated Adaptive Sleeping for Wireless Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The S-MAC protocol is a MAC protocol with wireless sensor networks in mind. In order to lower energy costs, S-MAC uses a coordinated adaptive sleeping technique to reduce energy usage and prolong overall life in a sensor network.
Each node in the network keeps track of its own schedules as well as its neighbor’s schedules. This schedule table defines the sleep periods for the nodes and based on this table, the nodes coordinate sleep among each other. Synchronization is an issue when coordinating schedules, but the paper maintains that clock drift is not a significant by correcting the drift using information from a SYNC packet and the packet transmission packet time.
The other technique in the S-MAC protocol is to use adaptive listening, so latency is reduced due to the sleep periods. A node will overhear a packet and temporarily wake-up to perform some type of processing before its scheduled time.
With sensor networks in mind, S-MAC is one of the alternative MAC protocols in development, but the adaptive listening portion seems counter-intuitive to one of the stated goals of avoiding overhearing packets. Aside from this detail, S-MAC’s basis will further research in wireless sensor networks.
- W. Ye, J. Heidemann and D. Estrin, "Medium Access Control with Coordinated, Adaptive Sleeping for Wireless Sensor Networks," IEEE/ACM Transactions on Networking, June 2004.
MACAW: A Media Access Protocol for Wireless LAN's
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
In response to another type of wireless LAN, the paper introduces the preliminary design of MACAW, which is an improvement of MACA. MACAW provides a shared knowledge of information by adding additional information in the packet headers, so other stations can make smart decisions when contending for the medium. This is in contrast to both CSMA and MACA, which both protocols use a grab-all approach for the medium.
MACAW proposes modifications to MACA that allows each station to have knowledge of other stations. Information such as congestion is important in order to share the media access, and contention periods synchronized across all devices so an overall picture can be established. In order to achieve these primary goals, MACAW looks in two areas of the MACA protocol, the RTS-CTS-DATA exchange and the binary exponential backoff algorithm.
The paper modifies the basic RTS-CTS-DATA exchange by adding extra packets, such as an ACK and DS. The ACK is used after each data packet transmission to ensure delivery of data and modifies the backoff counter appropriately. The DS packet is used to signify a successful RTS-CTS handshake and prepare for data sending.
The protocol is used for wireless LAN purposes but not specifically designed for low-power sensor networks. As stated by the paper, power considerations were not evaluated, and the paper does not include any type of duty cycling in the MACAW protocol to show evidence of power savings. It does provide a basis for other protocols since it does have better throughput with allocation to all streams instead of just having one dominate.
- V. Bharghavan, A. Demers, S. Shenker, and L. Zhang, "MACAW: A Media Access Protocol for Wireless LANs," SIGCOMM Symposium on Communications Architectures and Protocols (SIGCOMM'94), September 1994.
Ultra-Low Duty Cycle MAC with Scheduled Channel Polling
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Scheduled Channel Polling MAC (SCP-MAC) is another MAC protocol with sensor networks in mind. It has some roots with S-MAC, but it is based on the concept of channel polling with distributed schedules. The schedule scheme is similar to the coordinated efforts of S-MAC.
The channel polling method presented by the paper uses an adaptive method in order to see if the medium is available. This adaptive method allows it to adjust to variable traffic patterns. Polling is divided into slots and each node has a set number of regular slots for polling. The adaptive polling comes into play when the node analyzes its traffic and adds up to a number of dynamic polling slots to adjust to bursts of traffic.
The results presented by the paper show significant differences between SCP-MAC and low-power listening. Energy consumption in SCP-MAC is considerably reduced as the number of nodes increased. Throughput stays fairly consistent as the number of nodes increase for SCP-MAC.
Because it is scheduled, latency is an issue, but the adaptive polling reduces latency. As the number of nodes increases, adaptive polling shows significant improvements. Large problems do not stand out in the protocol as of now, but some problem areas might include synchronization issues with the scheduling especially determining network availability with a percentage of failed nodes.
- W. Ye, F. Silva, and J. Heidemann, "Ultra-Low Duty Cycle MAC with Scheduled Channel Polling," ACM Conference on Embedded Networked Sensor Systems (SenSys'06), November 2006.
The Tenet Architecture for Tiered Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper addresses the problem of unwieldy sensor network systems based on current data fusion principles. In order to alleviate this problem, the authors introduce a concept architecture known as Tenet. Tenet introduces using a tiered structure such that there is a lower tier and upper tier. The lower tier consists of motes while the upper tier consists of more powerful nodes. The main concept behind Tenet is only the upper tier performs data fusion and multi-node application logic. By this principle, motes only pass communication through the upper tier in order to communicate with other motes.
The Tenet architecture aims to provide a development framework for tiered network sensor systems and the results presented seem to show that it works, but more work is needed. Since tiered networks are aimed for large, scalable networks, the Tenet architecture needs to show it can meet that scalability in its results.
- O. Gnawali, B. Greenstein, K.-Y. Jang, A. Joki, J. Paek, M. Vieira, D. Estrin, R. Govindan, and E. Kohler, "The TENET Architecture for Tiered Sensor Networks," ACM Conference on Embedded Networked Sensor Systems (SenSys), November 2006.
SATIRE: A Software Architecture for Smart AtTIRE
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper addresses the need of mobile monitoring systems embedded in clothing in order to monitor human activity. The paper does not really introduce any new and exciting concepts that are directly related wireless sensor network issues. The same general frameworks are presented with some slight modifications to focus on embeddable monitoring sensors.
One interesting concept from this paper is activity detection based upon data from the accelerometers. By using the methods of presented in this paper to perform activity recognition with a full understanding of Hidden Markov Models, then the spread of this concept is can be widely applicable to sensor motes. The problem I see with this method is the need to perform profiling for an event before recognition can occur, which excludes attempting to identify anomalies that occur outside of the recognized spectrum.
- R. Ganti, P. Jayachandran, T.F. Abdelzaher, and J.A. Stankovic, "SATIRE: A Software Architecture for Smart AtTIRE," International Conference on Mobile Systems, Applications, and Services (MobiSys), June 2006.
TSAR: A Two Tier Sensor Storage Architecture Using Interval Skip Graphs
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper presents a way of exploiting tiered architectures to perform energy efficient storage of data and events. Previous methods employed flat hierarchical models of storage, which showed heavy usage of resources. This architecture, TSAR, aims to resolve that problem by separating data and metadata. In the tiered architecture, data and events would sit locally on motes. Metadata is moved into the upper tier on the more powerful nodes (proxy) in a distributed fashion among all the proxies.
In order to achieve the efficiency in storage, the metadata is placed into a data structure known as an Interval Skip Graph. The graph is distributed among all the proxy nodes and allows for range searches instead of single-value searches.
A primary issue is the loss of a proxy with a key piece of metadata for some requested data. The authors suggest adding redundancy into the distribution of metadata so another proxy will have the information if the primary proxy fails.
- P. Desnoyers, D. Ganesan and P. Shenoy, "TSAR: A Two Tier Storage Architecture Using Interval Skip Graphs", ACM Conference on Embedded Networked Sensor Systems (SenSys), November 2005.
VigilNet: An Integrated Sensor Network System for Energy-Efficient Surveillance
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
VigilNet is an energy-efficient integrated sensor network system for use in surveillance. The authors integrated various research ideas from literature pertaining to wireless sensor networks and created a prototype system, VigilNet. VigilNet is designed to track an object remotely and send the data back to a central location to provide intelligence of an area of interest.
The requirements set forth in the paper include longevity, adjustable sensitivity, stealthiness, and effectiveness. This paper introduces some key concepts in the areas of longevity and stealthiness. One of the assumptions in this paper is the deployment of nodes is fairly dense to provide redundancy as well as cycling through nodes to conserve power. In order for scalable deployment, VigilNet employs a hierarchical structure with the use of data aggregation to contribute to longevity. The active/non-active duty cycle is based on the concept of grouping motes within an area together and using a group of them at one time. During the cycle, groups of motes will alternate between active and non-active.
Stealthiness refers to limiting the amount of active transmissions so as not to be detected. In order to reduce transmissions, control messages for setup and data transmission are limited. Data aggregation is also utilized, so not every mote is sending data to a parent node. This also aids in reducing power. They implement data aggregation as a mapping of an event id and an event so as to create a unique one-one mapping in the groups of motes. This allows for the parent of the groups to transmit non-redundant data back to a base station and also reduces the number of transmissions.
The results presented by the paper show typical performance curves of various metrics to provide evidence that the system works as stated in the paper. Interestingly, data aggregation results show message overhead is reduced as the degree of aggregation is increased. Sleep cycle results show a fairly huge improvement with the usage of a reactive method than a proactive method to reduce the number of control messages in a network.
This paper provides a good example of how the pieces found in wireless sensor network literature fits together, and how wireless sensor networks truly perform when combined. The paper does not introduce any concepts that are amazing and it seems the surveillance systems is geared towards one hardware architecture. In order to port VigilNet to another platform, possible rework of the underlying code will need to be rewritten specific to the platform so ease of portability is not one of the factors.
- T. He, et al. "VigilNet: An Integrated Sensor Network System for Energy-Efficient Surveillance." ACM Transactions on Sensor Networks, 2(1): 1-38, February 2006.
Design and Deployment of Industrial Sensor Networks: Experiences from a Semiconductor Plant and the North Sea
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper introduces the concept of applying sensor networks to industrial situations for predictive maintenance of equipment. In order to illustrate the main points, the paper presents two scenarios of deployment, a semiconductor plant and an oil tanker. The overall architecture is the same between the two scenarios with minor differences in other details due to environmental layout.
The premise of the paper is to cost-effectively employ a proactive maintenance system through the usage of wireless sensor networks. The deployed motes are distributed in a hierarchical cluster fashion with heterogeneous nodes at each tier level. The industrial sensors were implemented on two different mote platforms, Mica2 and Intel Mote. The implementation of the industrial sensor network contains many concepts related to wireless sensor networks but also some of the concepts are not geared towards wireless sensor network architecture. 802.11 is used as part of the network backbone to provide timely communication. In order to counteract the effects of power consumption of 802.11, a cluster-based sleep/wakeup protocol was implemented.
During deployment, both the Mica2 and Intel Mote were compared. The Intel Mote is relatively higher in computing than the Mica2, and the paper’s results show the Intel Mote is more efficient in power consumption than the Mica2 overall.
Based on what is presented in this paper, the system takes some steps backwards in wireless sensor network development. Some constraints dictating wireless sensor network development to achieve long lifetime and optimized communication are important regardless. Though this is more cost efficient than the current methods, the implementation presented in this paper may not bode well as current research leans towards a more general unified wireless sensor network architecture.
- L. Krishnamurthy, R. Adler, P. Buonadonna, J. Chhabra, M. Flanigan, N. Kushalnagar, L. Nachman, and M. Yarvis, "Design and Deployment of Industrial Sensor Networks: Experiences from a Semiconductor Plant and the North Sea", ACM Conference on Embedded Networked Sensor Systems (SenSys), November 2005.
Vineyard Computing: Sensor Networks in Agriculture Production
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper is an analysis of wireless sensor network’s applicability to agricultural needs. The paper surveys various areas of agricultural that can benefit the use of wireless sensor networks.
One of the first issues introduced for proactive agriculture is monitoring irrigation needs. The paper states that water is a limited resource and dynamically, automated management of irrigation needs would increase efficiency and allow a vineyard manager/worker to allocate the time to other tasks. Pests detection falls within this same category though the paper discusses methods of detecting and then taking action on a pest.
In addition, the paper addresses the issue of data representation for the entire scope of a vineyard. The paper suggests moving towards the concept of human touchpoints and interacting different paradigms to be able to include the various different touchpoints within an organization.
Finally, the paper discusses some system architecture issues and states the ideal concept related to wireless sensor networks. One interesting concept that arises from the paper is the use of a data mule to provide a lower cost initiative in using wireless sensor networks for agriculture. The use of a data mule requires some mobile unit, whether an organic or non-organic host, to move the unit in order to collect data from the sensor nodes.
This paper presents some interesting issues related to agriculture that can be applied to wireless sensor networks in general. Adding a mobility factor for nodes to create a data mule can reduce equipment cost, so entry into using a wireless network sensor becomes more feasible. In the technical area, not much technical content is presented in this paper.
- J. Burrell, T. Brooke, and R. Beckwith, "Vineyard Computing: Sensor Networks in Agricultural Production", IEEE Pervasive Computing, pp. 38-45, 2004.
Wireless Mobile Ad-hoc Sensor Networks for Very Large Scale Cattle Monitoring
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper presents a solution to monitoring cattle livestock for diseases and other general heath issues to prevent outbreaks that might spread to the human population. The paper argues that a nationwide cattle monitoring system would decrease the number of potential outbreaks and be cost-efficient as well. The proposed system introduces a sensor system that is mounted on an animal, possibly the collar area. The system is organized as a hierarchical topology that will interface with the Internet for efficient data monitoring.
In order to provide an efficient monitoring system, the authors of the paper propose a solution in the system architecture. They propose to integrate the application layer, distributed hash tables (DHT), and a routing protocol for high efficiency. The claim for their reasoning is to build a robust DHT implementation that can withstand high disconnection rates and dropping success ratios since the environment is a highly mobile environment. The routing is basically the same as those found in Mobile Ad-hoc Network (MANET) specifications except with the added layer of DHT included.
The storage mechanism found in this system uses caching, and the paper classifies two types of methods, passive and proactive. Passive works in the state of promiscuous mode and will cache the information if a node is relatively close in terms of a special id number based on the extended routing protocol in this paper. Proactive involves exchanging information with single-hop neighbors based on the “freshness” of the state and node’s id relatively closeness to the communicating node’s id.
The evaluation of this storage/routing protocol shows fairly significant improvements over “normal” mechanisms. The proactive combined with passive caching shows significant improvements in terms of success ratio in the simulation environment presented by the paper.
The proactive mechanism presented in this paper adds some overhead in terms of communicating, which will affect energy consumption. Also, the paper paints a big picture of deploying a nationwide sensor network, but their simulation results deal with only one pasture of 100 nodes. The results show that a small scale works, but it does not show the effects of scaling the experiment to several orders of magnitude larger.
- M. Radenkovic and B. Wietrzyk, "Wireless Mobile Ad-hoc Sensor Networks for Very Large Scale Cattle Monitoring", Proc. of Sixth Int'l Workshop on Applications and Services in Wireless Networks (ASWN'06), Berlin, Germany, 2006, pp. 47-58
Monitoring Volcanic Eruptions with a Wireless Sensor Network
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper discusses the application of a wireless sensor network to monitor volcanic eruptions. The purpose of this paper is to provide an alternative to wired sensors that are currently in use in order to expand the usability of wireless sensors in almost any location regardless of geographic difficulties. The platform chosen is the Mica2 with a more powerful aggregation node. The network structure is a hierarchical structure to aid in streaming data back to a monitoring station. The implementation is typical as found in most proof-of-concept papers related to wireless sensor networks.
Some issues do arise in their implementation that affects the lifetime of deployed motes. As presented, the deployment was used in a continuous monitoring mode for 54 hours, which does not exactly reflect a power-conscious system. Possible workaround to this issue is to switch from a continuous operation to a threshold detection system though this depends on the needs of scientists in the seismology and volcanic areas. False positives are a possibility, but with adequate review of existing seismic data, false positives can be reduced if extended to a threshold detection system instead of a continuous operation.
- G. Werner-Allen, J. Johnson, M. Ruiz, J. Lees, and M. Welsh, "Monitoring Volcanic Eruptions with a Wireless Sensor Network," EWSN'05.
Chord: A Scalable Peer-to-Peer Lookup Protocol for Internet Applications
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Chord is a lookup protocol for peer-to-peer networks that addresses efficient location of a node for a preferred data item. The main features provided by Chord are scalability with handling for node connects/disconnects. In order to provide this feature, Chord utilizes a hashing algorithm and uses a key that will map onto a node. Where Chord differs is the view of network topology. The protocol requires that each node have a partial view of the topology since the routing table is distributed.
Digging deeper into the Chord protocol, 4 key areas define the simplicity of the Chord protocol. The 4 key areas are consistent hashing, simple key location, scalable key location, and dynamic operations/failures. The key guarantees that consistent hashing provides is a load-balanced distribution of keys. The remainder of the key areas provides robustness and efficiency in the protocol. The keys are distributed among a number of nodes and provide a distributed lookup so that each node does not need to know about every other node. The algorithms presented in scalable key location and dynamic operations/failures provide resiliency when a node joins/leaves because only a minimum number of keys are moved around to provide a load-balanced environment.
The paper addresses some key issues related to routing in sensor networks. The results presented in the paper along with the concepts show that the protocol delivers, and the algorithms seem fairly simple to implement. Some key issues still remain and have been at least addressed by the authors. Some of these issues include network partitioning and the insider threat of malicious nodes.
- I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan. "Chord: A Scalable Peer-to-Peer Lookup Service for Internet Applications." Technical Report TR-819, MIT, March 2001.
The Sensor Network as a Database
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper presents a concept architecture to speed application development for wireless sensor networks. The concept is analyzing wireless sensor networks as a database and applying the theories of databases to sensor networks along with wireless sensor network concepts.
In order to conceptualize a sensornet database architecture, the paper presents some key issues. The sensor network subsystems normally have a layered approach, whereas the paper suggests possibly breaking that approach to gain efficiency and robustness. Data models is another issue that incorporates describing the data. Finally, database operators such as join and aggregate must be tuned for sensor network operation to achieve efficiency.
The primary focus of the paper is on database operations, joins and aggregations. The paper discusses using non-blocking joins since data from a sensor network is unbounded through the use of pipelining and partitioning.
The use of aggregation is to introduce energy-efficiency and refine the data that gets passed around, so raw data is not sent back to the requestor. The paper proposes energy-efficient aggregation by stating that approximations work better than trying to determine an exact result. They introduce various approximating methods such as logarithmic sampling and hypothesis testing.
This paper primarily introduces a conceptual architecture that has no implementation yet based on the paper. The database architecture shows one way of architecting wireless sensor networks, but has some considerations of tight integration of the sensornet subsystems in order for efficiency to be achieved. This will probably help application developers, but might restrict a more open architecture for lower level design.
- R. Govindan, J. M. Hellerstein, W. Hong, S. Madden, M. Franklin, S. Shenker, "The Sensor Network as a Database," USC Technical Report No. 02-771, September 2002.
Beyond Average: Toward Sophisticated Sensing with Queries
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The overall goal as presented by this paper is to provide a simple framework for sensornet application developers that is simple and makes sense. The world behind almost every transaction or website is run by data, and the engine behind that is a database. The authors provide a high-level query language that is based on the TinyDB SQL language with their own added extensions.
Some new features they have added into the TinyDB SQL language is a mechanism to provide autonomic behavior for event detection. Provided some external stimulus triggers a sensor, an event can be stored and sent back to some central location. In addition to detecting events, the paper presents storage points. Storage points create a buffer to store data that is continually sampled, so if requested, the data can be processed through another query.
The paper further discusses the preference of having a multi-resolution histogram, but it also explains that it provides this at the expense of more communication. The work presented in this paper has good intentions, but it also has a feel that the authors are providing these features at the expense of power, which is a key component for sensor networks to be a really viable solution for wide use. The storage point concept is a continually sampled structure, and as stated by the authors, the multi-resolution histogram requires additional communication.
- J. M. Hellerstein, W. Hong, S. Madden, and K. Stanek, "Beyond Average: Towards Sophisticated Sensing with Queries", IPSN'03.
Synopsis Diffusion for Robust Aggregation in Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper addresses the problems of duplicate data traversing the network and data loss from losing a branch when the topology is in a tree structure. In order to reduce duplicates, synopsis diffusion utilizes in-network aggregation and applies its algorithm to summarize the data further to reduce duplicates.
In order to prove that the algorithm works accordingly, the authors apply synopsis diffusion to an adaptive ring topology though they claim topology independence. Adaptive Ring topology handles changes of various network changes such as node joins/leaves and link loss. It handles the changes by adapting the ring assignments of the nodes.
Two phases make up the synopsis diffusion algorithm: distribution phase and aggregation phase. The distribution phase floods an aggregate query throughout the network and constructs an aggregate topology. The aggregation phase continually routes data towards the querying nodes.
The paper uses formal methods to show that the synopsis diffusion algorithm is correct, but they also provide some realistic experimental results. Based on the data, the algorithm seems to be fairly robust based on the claims of the paper. The paper still does not address some key issues such as scalability of the algorithm. They do show node density as part of their results, but they do not discuss applicability of their algorithm when network sizes become excessively large.
- S. Nath, P. Gibbons, S. Seshan and Z. Anderson, "Synopsis Diffusion for Robust Aggregation in Sensor Networks," ACM SenSys, Baltimore, MD, November 2004
Minimum Power Configuration in Wireless Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Power consumption is one of the most important issues in wireless sensor networks. This paper proposes a solution for power management and conservation in wireless sensor networks. The solution provided in this paper shows that issues once thought isolated are in fact related. These issues include topology, routing, and sleep management. As such, the paper relates that a minimum power configuration is related to the data rate of sources.
The solution that is provided by the paper is essentially finding the route with the lowest cost in terms of power. A weighted graph is presented with power cost representing the weights. The paper introduces several algorithms and finally settles on using Incremental Shortest-path Tree Heuristic in order to find the path with the lowest cost in terms of power.
With these concepts in mind, the authors designed a minimum power configuration protocol (MPCP). MPCP is a routing protocol that is based on the destination-sequenced distance vector (DSDV) protocol. The routing table contains data rate information, next hop, cost, and sequence number for each link.
The simulation results presented in the paper show that MPCP performs relatively well in simulated conditions though MPCP does have a higher overhead in terms of control packets. In terms of power management, it outperformed the other 2 protocols that were simulated as well. One thing the paper does not mention is the active power consumption for processing the presented algorithms. Though radio transmission is an important consideration in terms of power usage, on-board processing is another area for power management and has been neglected in this paper.
- G. Xing, C. Lu, Y. Zhang, Q. Huang, and R. Pless, "Minimum Power Configuration in Wireless Sensor Networks," ACM International Symposium on Mobile Ad Hoc Networking and Computing (MobiHoc), May 2005.
Flexible Power Scheduling for Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Flexible Power Scheduling (FPS) is one of the many proposed solutions for an adaptive network power management for sensor networks. In this paper, the authors explain the natural characteristics of a time-slotted mechanism versus using a method that is deterministic in nature. The slotted mechanism allows for duty cycling the radio power in order to conserver overall energy in the motes.
The paper also included other features into FPS in order to make it more adaptive. The FPS protocol adapts to network fluctuations based on two levels. The paper describes that scheduling is done on two levels. Coarse-grained scheduling is performed at the routing layer to determine when the radio is on/off. Fine-grained scheduling is performed at the MAC layer to schedule channel access. They base the schedule assignment and modification on the concept of supply and demand. Supply refers to transmit slots while demand refers to receive slots in a node.
The algorithm starts that each node maintains its own power schedule. Time slots can be in one of 6 states as defined by the paper. A parent broadcast an advertisement to reserve a time slot and mark its own schedule in the receive pending state. A child with demand will accept the broadcast and mark its own schedule as transmit pending. The paper also reduces overhead by not having a requirement for a global initialization. Base station picks a random time slot and all nodes initiate with a demand one greater than their current demand.
The results show that FPS is reducing power overall. Their calculations from their simulation and experiment show that they drastically improved battery life from 9 days to 53 days. Ideally, the power source is infinite through some other means, but this drastic improvement in battery life shows that FPS works. In terms of message traffic, being the protocol is based on slots, messages eventually get to their destination with some latency. Based on the results, FPS seems like it is a worthwhile option to include in some applications. Latency may become a factor for critical applications, and the energy savings need to be looked at in conjunction with other areas of power conservation such as memory and processing.
- B. Hohlt, L. Doherty, E. Brewer, "Flexible Power Scheduling for Sensor Networks." Information Processing in Sensor Networks, 2004. IPSN 2004. Third International Symposium on, 26-27 April 2004, pp. 205-214
Twinkle: Network Power Scheduling in Sensor Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Twinkle is a new adaptation from the FPS paper. The authors included new extensions in order for FPS to be applicable to real-world applications. As stated in the paper, they added 4 new extensions: partial flows, broadcast, time sync, and latency optimizations.
Twinkle includes partial flow that allows for data aggregation, data compression, and query dissemination in the protocol. The paper states that broadcast is an instance of a partial flow. In reference to Twinkle, the partial flow is known as the Comm channel. The Comm channel is a broadcast channel used for transmitting synchronization packets and forwarding messages.
Twinkle also includes time synchronization from another protocol, flooding time synchronization protocol (FTSP). FTSP includes time-stamping in the MAC layer with skew compensation. The latency optimizations were performed in two areas of scheduling, reservation windows and fractional flows. The reservation window places an upper bound on per hop latency. Fractional flows divide slots into every k cycles instead of the normal one slot every cycle. This reduces the need for long cycles for infrequent slots.
Also in this paper, the authors showed two instances of real-world applications for Twinkle. They included Twinkle in a version of the Great Duck Island experiment and the Redwoods. Both experiments showed that Twinkle provided equivalent or better results in terms of power consumption with a duty cycle approach to power management. With the further work from FPS shown in this paper, the power management approach provided by Twinkle shows that a slot mechanism possibly works better in some monitoring applications.
- B.A. Hohlt and E.A. Brewer, "Twinkle: Network Power Scheduling in Sensor Networks," Technical Report No. UCB/CSD-05-1409, EECS Department, University of California, Berkeley, 2005.
Low-Energy Sensor Network Time Synchronization as an Emergent Property
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper addresses the need for a localized time synchronization method instead of a centralized broadcast mechanism. The introduction of pulse couple oscillations is the solution provided by the paper. The basic operation is that pulses are emitted and a time difference is determined between the last and current pulse. The time difference is then measured and normalized by some time interval, which will converge on some number between 0 and 1. 1 represents desired synchronization.
The simulations results show some promise in using a localized technique instead of a global time synchronization method. Power consumption is an important issue, but the figures presented in the paper are hardly decipherable. One figure did show promise was the clock variance related to node mobility. PCO showed great promise in this area based on the results. Though the paper was a little light in terms of effect on overall power, the PCO technique showed promise from some of the methods and results for a low-power localized time-synchronization technique.
- S.F. Bush, "Low-Energy Sensor Network Time Synchronization as an Emergent Property," Computer Communications and Networks, 2005. ICCCN 2005. Proceedings. 14th International Conference on. 17-19 Oct. 2005, pp. 93-98
The Semantic Web
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The article describes the next generation application of connecting services all at the whim of a request. The paper currently describes the semantic web as the next evolution of the Web. According to the paper, it is not a completely separate concept from the Web, but another layer built from the existing structure. In order to build this structure, the author suggests that data has a more defined representation but has the flexibility with diverse systems on a global scale. The author sees this achievement through the use of XML, RDF, ontologies, and agents to mesh it together. The largest focus will need to be on ontologies and agents because of varying theories in approach. The paper breaks down the various areas of focus for research needed to be able to build the semantic web. Though given such a high-level overview, each area has its own issues to work with, and eventually research in each area needs to merge with consideration in each area mentioned in this paper.
- T. Berners-Lee, J. Hendler and O. Lassila, "The Semantic Web," Scientific American, May 2001.
The Semantic Web Revisited
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper takes another look at the concept of the Semantic Web and its progress since Tim Berners-Lee introduced the concept approximately a decade ago from the time of publication of this paper. The paper surveys various projects that have some basis in forming parts of the Semantic Web. In particular, the paper focuses on standards and formalism for development because the Semantic Web’s goal is for global deployment and connectivity. Another focus of the paper is the development of ontologies, particularly in the scientific community. The paper shows a distinction between ontologies in the scientific community and the business community. The paper also goes into the cost factor for developing ontologies and states that the return on investment will far exceed the development cost. Another factor the paper considers is human nature. Human nature describes data in its own way through the use of tagging and is already in place as described by the paper. This paper provides a deeper look into the development of the Semantic Web and shows evidence of progress by naming a few projects out in the community already. The introduction of tagging from the perspective of a user as well as delving deeper into ontologies and standardization will lead to the goal of the Semantic Web.
- N. Shadbolt, T. Berners-Lee, and W. Hall, "The Semantic Web Revisited", IEEE Intelligent Systems, 21(3) pp. 96-101, May/June 2006.
OpenGIS Sensor Model Language (SensorML) Implementation Specification
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper is a proposal to define a standardized sensor language by defining various models and the base schema. SensorML is designed to describe processes, process chains, and sensors according to some fixed structure defined in the document. As with other papers previously seen, sensors are configured based upon manufacturer specifications. SensorML is another push to alleviate this narrow channel. The document goes through to define what is a sensor and what is its objective. It goes further to define the processes and process chains that a sensor is involved in and further defines a core schema based in XML. Defining a sensor based upon an XML structure is one way of modeling a sensor. Especially, XML has made its mark on the Internet for data delivery. Given this model, SensorML will be sure to make its mark in terms of providing a unified structured language for sensors.
- Open Geospatial Consortium, OpenGIS Sensor Model Language (SensorML) Implementation Specification, Document #OGC 05-086r2, Feb. 2006.
Plug-and-Play Sensors in Wireless Networks
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
This paper introduces a technology that is an example of the goal of IEEE 1451. The CrossNet network interfaces smart transducers to powerful machines such as handhelds or desktop computers. The paper continues to discuss the benefits of a standard interface scheme such as transducer datasheets, wiring, wireless networking. From these issues, the paper introduces the CrossNet architecture, specifically the CrossNet node. The CrossNet node is the bridge between a sensor and a more power system that will collect data through a wireless link. Bluetooth was chosen as the wireless link for this system. The idea of plug-and-play sensors is a phenomenal idea and the paper addresses the issue for the need of a plug-and-play setup, but the paper fails to mention the energy aspect in wireless sensor networks. The CrossNet architecture initially does not seem like an energy efficient architecture, especially since the paper fails to mention that detail.
- M. Dunbar, "Plug-and-Play Sensors in Wireless Networks," IEEE Instrumentation & Measurement Magazine, Vol.4, No.1, pp.19-23, March 2001.
IEEE 1451: A Standard in Support of Smart Transducer Networking
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper provides a general overview of IEEE 1451. The general overview summarizes the 4 models under IEEE 1451. IEEE 1451.1 aims to simplify interfacing to a networked transducer by providing a high-level abstraction that incorporates different network protocols. IEEE 1451.2 aims to standardize the interfacing of a transducer to a microprocessor through the use of a Transducer Electronic Data Sheet (TEDS), data format, and 10-wire interface. IEEE 1451.3 provides a multidrop distributed system with the use of a bus line between controller and modules. IEEE 1451.4 combines the concepts of IEEE 1451.1 and 1451.2 to have a mixed mode device that will monitor analog and transmit data digital to some host. The idea of standardizing an interface for transducers sets the industry in the right direction. The smart transducer standard will allow for faster time to market and possibly simplify new innovations since interfacing becomes standardized.
- K. Lee, "IEEE 1451: A Standard in Support of Smart Transducer Networking," IEEE Instr. and Meas. Tech. Conference, May 2000.
IEEE 1451 Standard and Frequency Output Sensors: How to Obtain a Broad-Based Industry Adoption?
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper introduces extensions for the IEEE 1451 standard. In particular, the paper provides extensions for frequency output sensors for interfacing with IEEE 1451. The main point of the paper is the introduction of the UFDC-1 module that is based on a simple architecture of a frequency-to-digital converter, memory for TEDS, and a bus interface that will accommodate 1451.2 or 1451.4. The paper does not provide much more information than making claims about the UFDC-1 architecture. Though the paper does introduce an extension for frequency output sensors, the lack of implementation and results plus survey of existing work fails to convince me about using the architecture.
- S. Y. Yurish, "IEEE 1451 Standard and Frequency Output Sensors: How to Obtain a Broad-Based Industry Adoption?," Sensors & Transducers Magazine, Vol. 59, Issue 9, September 2005, pp. 412-418.
Semantics-Enabled Framework for Knowledge Discovery from Earth Observation Data Archives
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
The paper introduces a framework for retrieving images based on semantics from Earth observation archives. The paper introduces the Intelligent Interactive Image Knowledge Retrieval (I3KR). The drive for this project is to make some sense of all the raw data collected for Earth observations. As with any semantic application, ontologies are at the core. I3KR uses a shared ontology layer to map domain information. Various methods are presented in the paper to extract distinguishing features from images for queries. The I3KR builds upon earlier described concepts of the semantic web in order to extract data from earth observation databases. I3KR shows that it is possible to build towards a semantic web with diverse data stores.
- S.S. Durbha and R.L. King, "Semantics-Enabled Framework for Knowledge Discovery from Earth Observation Data Archives, IEEE Trans. Geoscience and Remote Sensing, Vol. 43, no.11, pp.2563-2572, Nov. 2005.
Wikipedia: Ultra-wideband
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Ultra-wideband is a term for some piece of technology that has an operating bandwidth above 500 MHz or 25% of the center frequency as defined by the FCC. The FCC has deemed the use of ultra-wideband as unlicensed between 3.1 – 10.6 GHz. The applications that use ultra-wideband vary from small personal devices to larger devices such as radar or imaging systems. Based on the article, ultra-wideband provides a high resolution time-of-flight to determine distance between transmissions. One of the emerging technologies that is finding its way into mainstream technology is IEEE 802.15.4. One of the primary benefits to using ultra-wideband is the capability for high data rates. As more technology pushes wireless technology such as wireless sensor networks, the incorporation of ultra-wideband will increase.
Hierarchical Temporal Memory: Concepts, Theory, and Terminology
Reviewed by: Wes Huang (wh69 AT msstate DOT edu)
Hierarchical Temporal Memory (HTM) is a new method of incorporating learning and cognitive functions into a computer. The concept of HTM was derived from biology. Specifically, HTM models the neocortex. According to the paper, HTMs perform 4 primary tasks: discover causes, infer causes, predict, and direct behavior. Looking in parallel to biology, the brain performs similar actions that seemingly “just happen.” In the paper, arguments are made for the hierarchical and temporal structure in the design of HTMs. Hierarchy has this built-in feature of sharing as well as storage efficiency. The paper also states that a hierarchy is fairly representative of the real world. The temporal property adds another dimension to the learning process. The argument from the paper is time allows the HTM to correlate different patterns together with one common classification. The paper uses the watermelon as an example. The concepts presented in this paper are interesting and definitely a push in the forward direction. HTMs will probably show up in wireless sensor network sooner or later, but other issues related to wireless sensor networks must be resolved before an application like HTMs will be prominent.
Presentations
A Survey on Routing Protocols for Wireless Sensor Networks
Presented by: Wes Huang (wh69 AT msstate DOT edu)
Presentation slides located here.
- K. Akkaya and M. Younis, "A Survey of Routing Protocols in Wireless Sensor Networks," Elsevier Ad Hoc Network Journal, vol. 3, pp. 325-349, 2005.
Semantic Streams: A Framework for Composable Inference over Sensor Data
Presented by: Wes Huang (wh69 AT msstate DOT edu)
Presentation slides located here.
- K. Whitehouse, F. Zhao, and J. Liu, "Semantic Streams: a Framework for Composable Inference over Sensor Data," European Workshop on Wireless Sensor Networks (EWSN), February 2006.
Twinkle: Network Power Scheduling in Sensor Networks
Presented by: Wes Huang (wh69 AT msstate DOT edu)
Presentation slides located here.
- B.A. Hohlt and E.A. Brewer, "Twinkle: Network Power Scheduling in Sensor Networks," Technical Report No. UCB/CSD-05-1409, EECS Department, University of California, Berkeley, 2005.
- B. Hohlt, L. Doherty, E. Brewer, "Flexible Power Scheduling for Sensor Networks." Information Processing in Sensor Networks, 2004. IPSN 2004. Third International Symposium on, 26-27 April 2004, pp. 205-214.
