An energy-efficient adaptive DSC scheme for wireless sensor networks
Introduction
Wireless sensor networks (WSN) are becoming an enabling technology for implementing a large number of parallel-distributed systems that interact with the physical world [1]. A wide range of applications of sensor networks has already been envisioned, including geographical surveillance, reconnaissance, machine monitoring, and battlefield operations. Large scale and special application environments of WSN require the sensor applications to be highly energy and bandwidth efficient. To improve energy and bandwidth efficiency of WSN, wide research has been carried out in diverse areas such as wireless communications, network routing, and transport.
Recently, the application of distributed source coding (DSC) to remote source locating and tracking in WSN has received much attention. DSC was proposed and studied in 1970s, to compress correlated sources without intercommunications between the sources [2], [3]. It has, therefore, the potential to save bandwidth and energy for the applications of source locating and tracking in WSN, where multiple sensors may detect a target and send correlated readings to an information sink [4], [5]. DSC compresses multiple correlated readings in a distributed way, while reducing the consumption of network bandwidth and transmission power as well as the packet collision in the wireless networks. Although many data aggregation techniques have been proposed to compress identical or partially identical sensor readings transmitted through a sensor node in the transmission paths, data aggregate requires the support of a structure, maintenance of which is costive for sensor networks [6]. Moreover, different readings may be transmitted to the sink over different paths. DSC becomes, therefore, an important alternative approach to data compression in WSN.
Prior work on DSC in WSN has focused on information-theoretic aspects such as achievable rate-distortion regions [7], [8], although some existing work has addressed the construction of distributed source codes. Aaron et al. proposed a distributed source code construction by using turbo codes [9], Pradhan et al. constructed a DSC framework based on algebraic trellis codes [10], but both turbo codes and trellis codes are computation-complex and hardware-costly for sensor nodes. Other authors considered an energy-efficient DSC scheme with Lagrangrian cost function, also based on trellis codes [11]. Xiong et al. provided a nice overview over the fundamentals of DSC and the current state of the art in [5], [12]. They also investigated a distributed source code constructed with the parity-check matrix of a binary Hamming channel code. However, the high correlation among sensor readings is different from the correlation in Hamming channel code, which makes this code construction method not suitable for DSC in sensor networks. Sartipi et al. proposed a scheme for DSC using low-density parity-check (LDPC) codes [13]. The proposed decoding algorithm prevents error propagation and heavy damage to the sources. Chou et al. constructed a simple distributed source code to adaptively compress spatially and temporally correlated sensor readings [4]. However, as pointed out in [14], the proposed DSC scheme is not efficient in terms of coding efficiency. We proposed a practical and efficient random-binning based DSC scheme for the quantized sensor readings with full leverage of the correlations between the sensor readings [14]. The proposed DSC scheme achieves high coding efficiency while maintaining low signal distortion. Challenges remain, however, as the wireless bandwidth is changing from time to time and the total network bandwidth is shared by a large number of remote sensors [15]. It is therefore necessary and beneficial to adapt the DSC scheme to the changing network.
In this paper, we continue our previous work on the DSC scheme for WSN remote source estimation, in two aspects. Firstly, in addition to the performance analysis on estimated signal to distortion ratio (SDR), we quantitatively analyze the overall network energy consumption using the detailed power consumption model for wireless sensor communications proposed in [16]. Although energy consumption is critical for WSN, the performance of energy consumption for DSC schemes has not been analyzed in the existing literature. Secondly, we propose an adaptive control mechanism for the DSC scheme, to flexibly optimize either SDR or energy consumption performance by adapting the source coding and transmission parameters to the network conditions. Simulations are carried out to validate the efficiency of the proposed DSC scheme and adaptive control mechanism for both saving energy and improving the quality of source estimation.
The remainder of the paper is organized as follows. Section 2 formulates the problem and introduces a multi-mode power consumption model for wireless sensor communications. Section 3 proposes a basic transmission scheme. Its performances for source estimation quality and energy consumption are also analyzed. Section 4 introduces the random-binning based DSC scheme and proposes a novel adaptive DSC scheme. The SDR performance of the DSC scheme is analyzed and used to find the optimal coding and transmission parameters in the adaptive DSC scheme. In Section 5, we analyze the energy consumption performance of the DSC scheme. Energy consumption is used as a constraint in the SDR optimization process of the adaptive DSC scheme. Numerical results are presented and discussed in Section 6. Section 7 concludes the paper.
Section snippets
Problem assumption
In this paper, we assume that a sensor network is designed to monitor remote targets. For simplicity, we assume there is at most one active target at any time. Once the target becomes active, it is observed by the surrounding sensors in a sensing cell. The surrounding sensors generate observation signals and transmit the signals via other sensor nodes to a sink for further processing and taking corresponding actions. As the focus of this paper is on the distributed source coding and adaptive
Basic transmission scheme
As discussed in Section 2, in the basic transmission scheme the quantized sensor readings are directly transmitted to the sink without DSC encoding. Unlike the adaptive DSC scheme, which finds the optimal DSC and transmission parameters by interactions between the sink and the sensors, in the basic transmission scheme, we assume that the sink and the sensors always know the best quantization and transmission parameters in order to achieve the maximum SDR or minimum energy consumption without
Adaptive DSC scheme
In this section we briefly introduce the random-binning coding used in the DSC scheme, and then propose an adaptive DSC scheme in which the source coding and transmission parameters are adaptively controlled to achieve the best source estimation quality under the bandwidth and energy constraints.
Energy consumption of adaptive DSC scheme
In this section we analyze the energy consumption of the adaptive DSC scheme. Similar to the basic scheme, in the current design, only the energy consumed in the sensing cell and the relay nodes for the active and transient modes are considered. The energy consumed at the sink and at the sensors for sampling and quantization are not considered. Furthermore, the energy consumption for DSC encoding in the sensing cell is also neglected because of the low complexity of modulo operation [4]. The
Numerical results
Simulator is implemented in Matlab to evaluate the performance of the adaptive DSC scheme. For the reported results, each value is obtained by averaging over 30 simulations. For each simulation, the network is simulated for 30 min. As the basic transmission scheme is simple and robust, it is used as a benchmark to evaluate our proposed DSC scheme. The values of simulation parameters, used to obtain the numerical results, are summarized in Table 1.
In the simulations, the mean of the Gaussian
Conclusion
In this paper, we studied the problem of remote source estimation in WSN with application of a random-binning based DSC scheme. Compared to a basic transmission scheme, in the DSC scheme the correlated sensor readings are jointly encoded in a distributed way and sent to the sink for decoding and further processing. We modeled the signal to distortion ratio (SDR) performance of both the basic transmission scheme and the DSC scheme, as a function of observation noise, quantization, network
References (18)
- et al.
A distributed and adaptive signal processing approach to exploiting correlation in sensor networks
Ad Hoc Networks
(October 2004) - et al.
A survey on sensor networks
IEEE Commun. Mag.
(August 2002) - et al.
Noiseless coding of correlated information sources
IEEE Trans. Inf. Theory
(July 1973) - et al.
The rate-distortion function for source coding with side information at the decoder
IEEE Trans. Inf. Theory
(January 1976) - et al.
Distributed source coding for sensor networks
IEEE Signal Process. Mag.
(September 2004) - et al.
A hierarchical energy-efficient framework for data aggregation in wireless sensor networks
IEEE Trans. Veh. Technol.
(May 2006) The rate-distortion function for the quadratic Gaussian CEO problem
IEEE Trans. Inf. Theory
(1998)- P. Ishwar, R. Puri, S.S. Pradhan, K. Ramchandran, On rate-constrained estimation in unreliable sensor networks, in:...
- A. Aaron and B. Girod, Compression with side information using turbo codes, in: Proceedings of the IEEE Data...
Cited by (8)
Practical data compression in wireless sensor networks: A survey
2012, Journal of Network and Computer ApplicationsCitation Excerpt :We might infer from these results that the algorithm did not deal well with multiple data types per sensor node as presented in Section 3. Both Tang et al. (2007) and Chou and Petrovic (2003) techniques are practical in cases of single cluster topology as illustrated in Fig. 3. However, in order to extend these schemes to a more complex topology as shown in Fig. 15, it is obvious that the decoders (nodes number 1 in each cluster) need a low-complexity coding as well as encoders since both of them suffer from power constraints.
Lifetime maximized data gathering in wireless sensor networks using limited-order distributed source coding
2011, Signal ProcessingCitation Excerpt :This is distinct with the classic DSC, in which no complexity constraints are assumed at the DGN. Data gathering in WSNs based on DSC is considered for different objectives in the literature [13–18]. Specifically, an algorithm to maximize the lifetime of a WSN cluster is proposed in [13], where an optimal DSC for all nodes is assumed with no complexity constraints.
Power scheduling of distributed estimation in sensor networks with repetition coding
2010, Signal ProcessingCitation Excerpt :Wireless sensor networks (WSNs) comprise a large number of distributed nodes. Nodes in a WSN can cooperate to perform advanced signal processing tasks, such as distributed detection [1,2] and distributed estimation [3]. There are three typical topology strategies: centralized, Ad Hoc and clustered WSNs.
Analysis of PKF: A Communication Cost Reduction Scheme for Wireless Sensor Networks
2016, IEEE Transactions on Wireless CommunicationsLossy Compression for Wireless Seismic Data Acquisition
2016, IEEE Journal of Selected Topics in Applied Earth Observations and Remote SensingOptimal Radio-Mode Switching for Wireless Networked Control
2014, Lecture Notes in Control and Information Sciences