Cross Layer Protocol Suite for Correlated Data Gathering in Wireless Sensor Networks
Project Description
Cross-Layer Protocol Design, Error Control and Packet Size Optimization
The energy restrictions of the sensor nodes,
the reduced computational ability, the need for low cost design
and the simplified hardware assumptions necessitate re-visiting
the classical protocol stack used in networking. In this project
we explore cross-layer techniques in protocol design, error control
and packet size optimization.
Cross-Layer Protocol Design
The results show that XLM achieves significant energy savings with very food
reliability performance in a multi-hop wireless sensor network. Moreover, the latency performance
of XLM is also comparable to the state-of-the-art layered protocol stacks.
Cross-Layer Analysis for Error Control
- A cross-layer methodology for the analysis of forward
error correction (FEC) schemes and Automatic Repeat reQuest (ARQ) in WSNs is developed in
this project. The study includes the effects of multi-hop routing and the broadcast nature
of the wireless channel. The results of our analysis reveal that for certain FEC codes, decreasing
the transmit power (thereby increasing the hop length) decreases both the energy consumption and
the end-to-end latency subject to a target packet error rate compared to ARQ. Thus, FEC codes can be
regarded as an important candidate for delay sensitive traffic in WSNs. On the other hand, transmit
power control results in significant savings in energy consumption at the cost of increased latency.
Moreover, the cases where ARQ outperforms FEC codes are indicated for various end-to-end distance and target PER values.
- This study is further extended to the case of hybrid ARQs (HARQs).
Mainly, two types of HARQ schemes exist: Type I and Type II. With HARQ-I techniques, first an uncoded
packet or a packet coded with a lower error correction capability is sent. If this packet is received
in error, the receiver sends a negative acknowledgement (NACK) to the sender, which re-sends the packet
coded with a more powerful FEC code. The difference in Type II is that for retransmissions, only the redundant
bits are sent. While Type II decreases the bandwidth usage of the protocol, Type I does not require the previously
sent packets be stored. The key insight gained in this study is that HARQ-I codes are slightly inefficient in terms
of both energy consumption and latency. HARQ-II scheme is more energy efficient compared to ARQ and some other comparable
block codes.
Cross-Layer Packet Optimization for Terrestrial, Underwater and Underground Channels
- For terrestrial sensor networks, the effect of packet length on the collision probability
is investigated. Moreover, the relationship between routing decisions and the packet size is highlighted in this project.
Furthermore, the effects of packet size on different performance metrics such as throughput, energy consumption, latency,
and success rate are investigated considering these cross-layer effects. Finally, requirements of various types of applications
in WSN are considered to develop a comprehensive framework for packet size optimization.
- The channel model for underground wireless communication developed at the BWN lab is chosen for
the study. Here, the path loss in an underground environment is a function of 1) the attenuation constant, 2) the volumetric water
content (VWC) of the soil, 3) bulk density, as well as 4) the mass fractions of sand and clay. We observe that the increase in
volumetric water content results in higher packet sizes for the energy consumption minimization problem, where the optimum energy
consumption also increases for higher values of volumetric water content.
- Underwater Acoustic Sensor Networks (UW-ASN) are characterized by the acoustic communication channel,
which shows an attenuation with frequency, besides very low propagation times. Interestingly, simulation results reveal that ARQ
schemes fare better than FEC in both shallow and deep water conditions and also need smaller packet sizes.
Publications:
- I. F. Akyildiz, M. C. Vuran and O.B. Akan, "A Cross Layer Protocol for Wireless Sensor Networks," in Proc. Conference on Information Sciences and Systems (CISS '06), Princeton, NJ, March 2006.
- M. C. Vuran and O. B. Akan, "Spatio-Temporal Characteristics of Point and Field Sources in Wireless Sensor Networks," IEEE Int. Conference on Communications (ICC), Istanbul, Turkey, June 2006.
- M. C. Vuran and I. F. Akyildiz, "Cross Layer Analysis of Error Control in Wireless Sensor Networks," Proc. of IEEE SECON, Reston, VA, September 25-28, 2006.
- M. C. Vuran, and I. F. Akyildiz, "Error Control in Wireless Sensor Networks: A Cross Layer Analysis," submitted for journal publication, August 2007.
- M. C. Vuran, and I. F. Akyildiz,"XLM: Cross Layer Module for Efficient Communication in Wireless Sensor Networks," submitted for journal publication, October 2007.
- M. C. Vuran and I. F. Akyildiz, "Cross Layer Packet Optimization for Wireless Terrestrial, Underwater and Underground Sensor Networks," in IEEE Infocom'08 Conf., April 2008
[Back to top]
MAC-Free Reading of Correlated Sensor Networks
The data gathering process is the first step towards realizing a complete network solution for WSNs. While
the classical approach establishes multihop routes to the sink from the sources, it involves considerable energy expense. In this
project, we investigate a cooperative approach for aerial reading of a wireless sensor network. More specifically, a data aggregation
method called as cooperative spectrum fusion (CSF) is devised to read data from the WSN without using Medium Access Control (MAC)
signalling.
Figure 1. Cognitive radio network architecture.
The aerial receiver (an unmanned or manned air vehicle) triggers a group of sensors using a beacon signal
transmitted by its directional antenna (Figure 2). The ground sensors respond to the beacon by offsetting the beacon signal
carrier frequency using their measurement of the field contour at their position. This is done simultaneously by all sensors.
The aerial node receives a superposition of these signals, i.e. a fused signal, and estimates the field contour using the
received signal. This approach eliminates the MAC layer as the sensors cooperatively respond in the same channel and since
the sensors respond at once, CSF also provides a fast way of reading a wireless sensor network.
Key Features and Results
- The received signal is a harmonical random process that is a sum of a number of complex exponentials. First, a simple way of using the spectral average of the received signal for estimation was proposed and tested through Monte-Carlo simulations. These simulations addressed different cases of network density and field correlation levels.
- A new sensor transceiver energy model was devised to reflect the latest hardware developments. The proposed model has a linear formulation of the dependence of transmitter energy cost on a radiated output energy. Furthermore, it was extended by inspecting a commercial node transceiver and identifying the model parameters in terms of its actual device parameters.
- The CSF approach is also shown to use less energy than conventional cluster-based, multi-hop or direct transmission protocols. The energy savings are in the order of minimum 50-90% savings for the earlier and the newly-proposed experimental transceiver configurations.
Publications:
- A. Akanser and M. A. Ingram, "MAC-free reading of a network of correlated sensors," Proc. IEEE Conference on Military Communications (MILCOM), 2007.
- A. Akanser and M. A. Ingram, "MAC-free Cooperative Spectrum Fusion (CSF) in Wireless Sensor Networks (WSN)," submitted to the IEEE Transactions on Aerospace and Electronic Systems (AES), 2008.
[Back to top]
Distributed Source Coding in Sensor Networks
We focus on two different types of correlation in this project - near and far correlation. We define near correlation as the correlation between content sent by sensors in the same vicinity. For example, the detection of the same event by multiple sensors in a region will result in near correlation. Similarly, we define far correlation as the correlation between content sent by sensors that are far apart. Such correlation can happen due to either large range events, or semantic correlation between data sent by far apart sensors.
We argue in this work that near and far correlation can be, and should be, handled differently in terms of the aggregation process for optimal performance. Specifically, in the case of near correlation, it is easier to predict in advance the existence of correlation between data sent by sensors in the vicinity, and hence a more aggressive aggregation scheme can be employed. At the same time, we argue that such prediction is infeasible for far correlation, and hence alternative strategies need to be employed. In this context, we consider two different, but complimentary, schemes for aggregating correlated data in wireless sensor networks that specifically target near and far term correlation respectively. The two schemes broadly fall under the classification of distributed source coding (DSC) and data gathering techniques.
We first assume that we have full knowledge of the correlation values before sensor deployment. This prior information is used to optimally design the source coding. Each node compresses its data without communicating with the other node and sends the compressed data to the next node that is closer to the sink. We propose a scheme for distributed source coding of correlated signals of two nodes. We show that our approach reaches the Slepian-Wolf limit, if we use a channel code that achieves the capacity of the equivalent channel. Next, we propose a method for compression of two correlated sources at every arbitrary rate on the Slepian-Wolf rate region using a single channel code. To describe the procedure, first we assume that each source uses a separate systematic LDPC code. Then, we show how the same procedure compresses both sources at rates close to the theoretical limit using only a single systematic channel code. We also extend our results for two sources to three sources by source coding of two signals together and the third one is compressed with the rate as close as possible to the theoretical limit. Assuming a large number of sensor nodes randomly deployed in a circular гд??eld, we propose a novel clustering scheme called Annular Slicing-based Clustering, and show that the proposed scheme performs near-optimally. Thus, we demonstrate that a judicious choice of the cluster size and distribution could result in better energy effciency, which is very valuable to the design of distributed WSNs.
Key Features and Results
- We studied the lossy distributed source coding which provides useful insight into interaction between source and channel coding. We found as to how we can achieve the Wyner-Ziv theoretical limit using practical error control techniques.
- Simulation results reveal that our scheme performs $0.2$ bits away from the Wyner-Ziv theoretical limit for the LDPC code of length 952. For longer LDPC codes oflength 1905, the gap from the theoretical limit decreases to $0.18$. This suggests that if longer LDPC codes are used, the performance would approach the theoretical limit.
Publications:
- M. Sartipi and F. Fekri, "Distributed Source Coding in Wireless Sensor Networks using LDPC Coding: A Non-uniform Framework," Proc. of IEEE Data Compression Conference, pp. 477-477, March 2005.
- M. Sartipi and F. Fekri, "Distributed source coding in wireless sensor networks using LDPC coding: The Entire Slepian-Wolf Rate Region," Proc. IEEE Wireless Communications and Networking Conference, pp. 1939-1944, March 2005.
- M. Sartipi and F. Fekri, "Distributed Source Coding using Finite-Length Rate-Compatible LDPC Codes: The Entire Slepian-Wolf Rate Region," IEEE Transactions on Commmunications, Vol. 56, No. 3, pp. 400--411, March 2008.
- M. Sartipi and F. Fekri, "Lossy Distributed Source Coding Using LDPC Codes," IEEE Communication Letters, Submitted January 2008, Revised and Resubmitted May 2008.
- B. N. Vellambi and F. Fekri, "Finite-Length Rate-Compatible LDPC Codes: A Novel Puncturing Scheme," IEEE Transactions on Commmunications, accepted, April 2008.
- R. Subramanian and F. Fekri, "A Clustering-based Framework for Energy Aware Data Gathering in Distributed Sensor Networks," Journal of Ad-Hoc and Sensor Wireless Networks, submitted, August 2008.
[Back to top]
Energy-efficient Data Gathering in Wireless Sensor Networks
We consider the problem of data gathering in environments where data
from the different sensors are correlated. In this project, we explore how best the data may be fused inside
the network using 1) cues from the network state, 2) energy-efficient aggregation trees rooted at the sink,
and 3) congestion reduction techniques.
- Cue-based Networking: This project presents a new approach called cue-based
networking that uses hints or cues about the physical environment to
optimize networked application behavior. We identify both the research
and system challenges, specifically the timeliness-robustness tradeoff,
that needs to be addressed to realize benefits of the approach under a
target application of video delivery over IP networks. We design an adaptive
algorithm that balances this tradeoff and test its performance through an
implementation of a video delivery application in a real home environment.
Results reveal that our prosposed probabilistic algorithm matches the benefits
sof continuous reporting, while maintaining a normalized reiability index of 1.
- Sink-to-sensor Congestion Control: Here, we focus on
providing congestion control from the sink to the sensors in a sensor field.
We identify the different reasons for congestion from the sink to the sensors
and show the uniqueness of the problem in sensor network environments. We propose
a generic framework that addresses congestion from the sink to the sensors in a
sensor network. We then propose an adaptive, explicit rate control approach,
called CONgestion control from SInk to SEnsors (CONSISE), that adjusts the
downstream sending rate at each of the sensor nodes to utilize the available network
bandwidth depending on the congestion level in the local environment. Simulation
studies reveal that CONSISE is able to mitigate the effects of congestion significantly
better for all multiple sinks by adjusting the sending rate to the available bandwidth,
incurring minimal loses.
- Scalable Correlation-Aware Aggregation: Sensors-to-sink data in WSNs
are typically characterized by correlation along the spatial, semantic, and/or temporal
dimensions. Here, we first identify that most of the existing upstream routing approaches
in WSNs can be translated to a correlation-unaware data aggregation structure - the shortest-path tree (SCT).
Although by using a shortest-path tree, some implicit benefits due to correlation are possible,
we show that explicitly constructing a correlation-aware structure can result in considerable performance improvement.
Toward this end, we present a simple, scalable and distributed correlation-aware aggregation structure that addresses
the practical challenges in the context of aggregation in WSNs. When compared with Decentralized Shortest Path Trees (
DPST), we observe that the cost of DSPT is up to 200\% of SCT cost, as the number of nodes increases. The cost of DSPT
s also increases faster than that of the SCT approach as node number increases.
Publications:
- S.-J. Park, Y. Zhu, R. Vedantham and R. Sivakumar, "A scalable correlation aware aggregation strategy for wireless sensor networks," in IEEE International Conference on Wireless Internet (WICON), Budapest, Hungary, 2005.
- K. Sundaresan, Y. Zhu, and R. Sivakumar, "Practical limits on achievable energy improvements and useable delay tolerance in correlation aware data gathering in wireless sensor networks," in IEEE Communications Society Conference on Sensor and Ad Hoc Communications and Network (SECON), Santa Clara, California, 2005.
- K. Sundaresan, Y. Zhu, and R. Sivakumar, "Exposing two critical myths about correlation aware data aggregation," in Poster Presentation, ACM International Symposium on Mobile Ad Hoc Networking and Computing (MOBIHOC), Urbana-Champaign, IL, USA, May 2005.
- R. Vedantham, R. Sivakumar and S.-J. Park, "Sink-to-Sensors Congestion Control," Elsevier Ad Hoc Networks Journal, vol. 5, no. 4, pp. 462-485, May 2007.
- Y. Zhu, R. Vedantham, S.-J. Park and R. Sivakumar, "A Scalable Correlation Aware Aggregation Strategy for Wireless Sensor Networks," Elsevier Information Fusion Journal, 2007.
- Y. Jeong, S. Lakshmanan, S. Kakumanu, and R. Sivakumar, "Cue-based Networking using Wireless Sensor Networks: A Video-over-IP Application," IEEE Comm, Society Conf. on Sensor, Mesh and Ad hoc Communications and Networks (SECON), San Francisco, CA, June 16-20, 2008.
[Back to top]
