A Survey on Machine Learning Techniques in Wireless Sensor Networks

Aziz Nasridinov, Young-Ho Park

Department of Multimedia Science, Sookmyung Women’s University,

Seoul, South Korea {aziz, yhpark}@sm.ac.kr

Abstract. Potential worldwide deployment of WSNs for, e.g., environmental monitoring purposes could yield data in amounts of petabytes each year. Thus, in addition to the highly interesting technical challenges related to WSNs themselves, their widespread deployment would also require development of solutions for analyzing the potentially huge amounts of data they would generate. In this paper, we present a survey on machine learning techniques in WSNs. We categorize machine learning techniques in WSNs into two types, such as machine learning techniques for networking and machine learning techniques for data processing.

Keywords: machine learning, data processing, energy-awareness.

1 Introduction

Most of the recent research in the field of wireless sensor networks (WSNs) has mainly focused on the technical challenges related to WSNs, such as developing solutions for energy efficient deployment, routing and management of WSNs. However, potential worldwide deployment of WSNs for, e.g., environmental monitoring purposes could yield data in amounts of petabytes each year [1]. Thus, in addition to the highly interesting technical challenges related to WSNs themselves, their widespread deployment would also require development of solutions for analyzing the potentially huge amounts of data they would generate.

Machine learning methods have been central in developing WSN applications since the very early days of WSNs, as many of the problems in WSNs could be put as optimization or modeling problems. Machine learning in WSN helps to discover meaningful new correlations, patterns and trends, often previously unknown, by sifting through large amounts of data, using pattern recognition, statistical and mathematical techniques. It can be useful not only in knowledge discovery, that is, the identification of new phenomena, but also it can help in enhancing our understanding of known phenomena. In other words, machine learning techniques can help build decision-aid tools and facilitate analyzing of sensor data obtained from WSNs.

In this paper, we present a survey on machine learning techniques in WSNs. We categorize machine learning techniques in WSNs into two types, such as machine

Advanced Science and Technology Letters Vol.30 (ICCA 2013), pp.106-108

http://dx.doi.org/10.14257/astl.2013.30.22

ISSN: 2287-1233 ASTL Copyright © 2013 SERSC

learning techniques for networking and machine learning techniques for data processing.

2 Machine Learning in Wireless Sensor Networks

We categorize machine learning techniques in WSNs into two types, such as machine learning techniques for networking and machine learning techniques for data processing. The rest of this section describes these techniques.

2.1 Machine Learning Techniques for Networking

Wang et al. [2] proposed a supervised learning approach for routing optimizations in WSNs. First, they develop a framework that uses supervised learning to automatically extract useful information within sensor networks. Second, they cast the link quality estimation problem as a classification problem, which permits the use of standard, yet effective algorithms. Third, they show that tree-based routing topologies in data collection applications may suffer from information loss, such as neighbor link quality, in an overloaded network. The authors used supervised learning to establish data collection trees in such adverse conditions.

Yusuf and Haider [3] present a novel fuzzy model for energy-aware routing in WSNs. The proposed method uses the disadvantage of not being easily adaptive to changes in sensor types because energy metrics vary widely with the type of sensor node implementation platform and some of the factors for calculating routing metric are conflicting. Authors argue that fuzzy logic, on the other hand, has potential for dealing with conflicting situations and imprecision in data using heuristic human reasoning without needing complex mathematical modeling. Thus, the method presents a soft, tunable parameter based approach using fuzzy variables and rule base. This results in a soft, accommodating routing protocol capable of catering to a wide range of energy metrics of different implementation platform.

2.2 Machine Learning Techniques for Data Processing

Careful selection of the aggregator nodes in the data aggregation process results in reducing large amounts of communication traffic in the WSN [4]. Thus, there are many approaches to achieve optimal selection of aggregator nodes.

Gupta et al. [5] a fuzzy logic approach to cluster-head election is proposed based on three descriptors - energy, concentration and centrality. Simulation shows that depending upon network configuration a substantial increase in network lifetime can be accomplished as compared to probabilistically selecting the nodes as cluster-heads using only local information. For a cluster, the node elected by the base station is the node having the maximum chance to become the cluster-head using three fuzzy descriptors - node concentration, energy level in each node and node centrality with respect to the entire cluster, minimizing energy consumption for all nodes consequently increasing the lifetime of the network.

Advanced Science and Technology Letters Vol.30 (ICCA 2013)

Copyright © 2013 SERSC 107

Nasridinov et al. [6] investigated optimal aggregator node selection in WSNs. The selection process is formulated as a top-k query problem, which is solved by adapting a modified Soft-Filter-Skyline (SFS) algorithm. The main idea in this approach is to perform a skyline query on the sensor nodes in WSNs in order to extract among those sensor nodes that are potential candidates for the leading role, and those that cannot possibly become an aggregator node. The approach selects a set of leading aggregator nodes according to their attributes, such as such as distance from the base station, power consumption, battery life and communication cost. Experiments show that our method outperforms existing approaches by up to several times in many cases.

3 Conclusion

In this paper, we have presented a survey on machine learning techniques in WSNs. We have categorized machine learning techniques in WSNs into two types, such as machine learning techniques for networking and machine learning techniques for data processing.

Acknowledgement. is work was supported by the IT R&D program of MKE/KEIT. [10041854, Development of a smart home service platform with real-time danger prediction and prevention for safety residential environments].

References

1. Jardak, C., Riihijärvi, J., Oldewurtel, F., Mähönen, P.: Parallel processing of data from very large-scale wireless sensor networks. In: Proceedings of the 19th ACM International Symposium on High Performance Distributed Computing, pp. 787--794 Chicago, Illinois, USA (2010)

2. Wang, Y., Martonosi, M., Peh, L. S.: A supervised learning approach for routing optimizations in wireless sensor networks. In: Proceedings of the 2nd international workshop on Multi-hop ad hoc networks: from theory to reality, pp. 79--86 (2006)

3. Yusuf, M., Haider, T.: Energy-Aware Fuzzy Routing for Wireless Sensor Networks. In: Proceeding of 2005 International Conference on Emerging Technologies, pp. 63--69 (2005)

4. Subhlok, J., Lieu, P., Lowekamp, B.: Automatic node selection for high performance applications on networks. In: Proceedings of the seventh ACM SIGPLAN symposium on Principles and practice of parallel programming, pp. 163--172, Atlanta, Georgia, USA (1999)

5. Gupta, I., Riordan, D., Sampalli, S.: Cluster-Head Election Using Fuzzy Logic for Wireless Sensor Networks. In: Proceedings of the 3rd Annual Communication Networks and Services Research Conference, pp. 255--260 (2005)

6. Nasridinov, A., Ihm, S. Y., Park, Y. H.: Skyline Based Aggregator Node Selection in Wireless Sensor Networks, International Journal of Distributed Sensor Networks (2013).

Advanced Science and Technology Letters Vol.30 (ICCA 2013)

108 Copyright © 2013 SERSC