Energy efficient data collection in sink-centric wireless sensor networks: A cluster-ring approach☆
Introduction
Internet of Things (IoT) sheds new light on the research of Wireless Sensor Networks (WSN) by expanding the notion of sensor node to include home appliances, surveillance devices, etc. Many IoT applications require continuous or periodic collection of sensing data in order to obtain a holistic view of network. For example, continuous data collection enables pre-cautioning for environment pollution tracking scheme in [18]. Traffic surveillance CCTV [16], pollution detector [15], fire detector [17], and video sensing [19], and agricultural sensing [20] are other examples.
Due to the high cost of cellular communication, multi-hop communication by using short range radio is still a preferable solution, particularly for the large scale WSN. The `cost' means not only the financial one but also the energy resources that is usually limited in WSN. As the network size is increasing, the entire size of collected data and number of node grows geometrically. This makes it harder to handle the limited energy resource of sensor nodes than the case of smaller network. Moreover, in multi-hop WSN, generated sensing data is often relayed to the sink node by multi-hop manner to get the holistic view of network. If the battery of a sensor node is depleted, not only its sensing operation but also data relaying is interrupted. By progression of energy depletion of sensor nodes, some node permanently loses all routes to the sink node, which is well-known as a network partitioning [12], [13]. In particular, the sensor nodes near the sink node are inevitably suffering for the high energy consumption in this sink-centric traffic pattern, which is known as `hot spot' problem. As the network size is increasing, this hot spot problem is exacerbated by the rapid increase of relayed data. To sustain the operation of network as long as possible, it is crucial to mitigate the hot spot problem.
In this paper, we suggest a solution to prevent network partitioning in a view of network organization and routing of data. We start from the idea that the circularly symmetric network is easy to be abstracted in a view of routing optimization. If the network is already formed circular-symmetrically, the energy consumption of network also can be easily controlled to be circular-symmetric, which gives a good abstraction rather than each area has different energy consumptions. By this idea, we first present an algorithm to compose a circular-symmetric cluster structure. For the given size of clusters, this algorithm gradually forms a cluster structure without requiring any location information of sensor nodes. We then present a pair of algorithms to decide traffic routing over the composed cluster structure. Existing cluster routing schemes commonly treat a cluster as the basic unit of the routing decision. In contrast, we group the clusters that is located with the same distance from the sink are equal. Since we already formed the network circularly symmetric, the grouped clusters form a doughnut-shaped `ring' of clusters. We call this group as `cluster-ring' or `ring' in this paper. If we assume that all clusters in the same cluster-ring does same actions for data relaying, the energy consumption pattern also becomes circularly symmetric. In other words, all clusters at the same distance from the sink node should have identical energy consumption.1 We divide the entire network into the set of cluster-rings.
With this abstraction, we convert the routing problem into two sub problems: (i) the decision of the amount of data flow between cluster-rings, and (ii) the decision of traffic relaying between clusters. Once we solve first sub problem, we solve the second problem on the basis of the solution of the first sub problem. The main benefit of the two-level approach is the scalability. If the number of clusters is very large, existing schemes that make routing decisions at the level of individual cluster will become quickly infeasible. By grouping the clusters into the `cluster-ring', the computational complexity can be significantly reduced.
For the ring-level routing (1st sub problem), we present an algorithm that yields energy-efficient strategy by using the reinforcement learning technique [7]. For the cluster-level routing (2nd sub problem), we present an algorithm for choosing the relay nodes for inter-cluster communication. The energy cost metric proposed in [8] is used to in this algorithm. Our two-level routing algorithms adapt to the dynamic changes of network topology and traffic pattern. The simulation results indicate that our scheme produces near optimal performance.
The rest of paper is organized as follows. Section 2 describes the related works. The network model and assumptions are presented in Section 3. In Section 4, we suggest a method to compose a cluster-ringed network. In Section 5, we formulate the problem in linear programming and propose a method that adaptively decides data flow between cluster-rings. Section 6 presents a cluster-level routing algorithm. The simulation results are shown in Section 7. Section 8 concludes the paper.
Section snippets
Related work
Energy-efficient routing for WSN has been extensively studied, mostly focusing on balancing the energy consumption of sensor nodes [23], [24]. Existing schemes oftentimes assume static traffic generation pattern and rely on the optimization solutions. In [4], a routing algorithm that maximizes energy efficiency based on Pareto optimality is proposed. This scheme is suitable for the case that the link configuration and traffic generation pattern can be probabilistically characterized. In [3],
Network model
While the proposed scheme can be applied to generic configurations, we make some assumptions on the network topology for the ease of presentation. That is, we assume that the network has a circular shape and the sink node locates at the center. Each sensor periodically generates sensing data at a certain rate while different sensor nodes may produce data at different rates. We consider a sink-centric traffic pattern, meaning that all the sensing data are forwarded to the (single) sink node.
The
Composition of cluster structure
Cluster structure is formed in two phases: (i) ring border decision, (ii) cluster border decision. In the ring border decision phase, the border of each ring is determined incrementally from the 1st ring. In the cluster border decision phase, the border between the clusters within each cluster-ring is determined. We do not require any location information such as GPS coordinates to this end.
Cluster-ring level routing decision
After forming the cluster structure, all nodes in network start generating data. Those data are ultimately collected to the sink node. To achieve this data collection, each node forwards its data to its CH and all CH forwards its collected data to the other CHs until those data is reached to the sink node. We define a `cycle' as a single execution of above process that starts from the data generation and ends in the arrival of all data to the sink node.
To maximize network lifetime, the route of
Cluster-level routing decision
In this section, we address the issue of the cluster-level routing decision after the ring-level routing decision is made. Ring-level routing decides how much traffic should be forwarded from which ring to which ring. Recall that cluster ring is conceptual and actual inter-cluster communication is the responsibility of `relay nodes' of the clusters. Each flow between two clusters involves two relay nodes: a sender-side relay and a receiver-side relay. To prevent the high energy burden of the
Simulation results
In this section, we evaluate the performance of the proposed scheme (i.e., einforcement learning-based algorithm). The solution of the LP model presented in Section 5.1 is used as the performance upper bound (i.e., the optimal solution). We firstly show that the result of the proposed scheme nearly converges to the optimal (LP) solution. We also show the scalability of the proposed algorithm for large size networks. We compare the performance of the proposed scheme with some existing works.
We
Conclusion
In this paper, we proposed a scalable scheme for energy-efficient data flow control for large-scale WSN. We start from the idea of network abstraction by forming the circular-symmetric clustered network. The main novelty of our algorithm is the division of the routing problem into two sub problems by adopting the cluster structure. Such division allows us to achieve scalability and to apply heterogeneous approaches to each problem. We used the reinforcement learning technique for the
Acknowledgements
This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIP) (No. 2016R1A2B4014505).
Soo-Hoon Moon received the BS and MS degrees in computer science from Yonsei University, Korea, in 2008 and 2010, respectively. Since 2010, he has been pursuing his Ph.D degree in the Department of Computer Science, Yonsei University, Seoul, Korea. His current research interests include sensor networks, green networking and embedded systems.
References (36)
- et al.
Wireless sensor network-based fire detection, alarming, monitoring and prevention system for Bord-and-Pillar coal mines
J. Syst. Softw.
(2012) - et al.
A novel methodology for the monitoring of the agricultural production process based on wireless sensor networks
Comput. Electron. Agric.
(2011) - et al.
Energy efficient and QoS based routing protocol for wireless sensor networks
J. Parallel Distrib. Comput.
(2010) - et al.
An energy-balancing clustering approach for gradient-based routing in wireless sensor networks
Comput. Commun.
(2012) - et al.
An energy-efficient and sink-location privacy enhanced scheme for WSNs through ring based routing
J. Parallel Distrib. Comput.
(July 2015) Multi-agent reinforcement learning: Independent vs. cooperative agents
- et al.
A survey on clustering algorithms for wireless sensor networks
Comput. Commun.
(2007) - et al.
A survey on routing protocols for wireless sensor networks
Ad Hoc Netw.
(2005) - et al.
COCA: constructing optimal clustering architecture to maximize sensor network lifetime
Comput. Commun.
(2013) - et al.
Prolonging the lifetime of wireless sensor networks via unequal clustering
HYMN: a novel hybrid multi-hop routing algorithm to improve the longevity of WSNs
IEEE Trans. Wireless Commun.
Energy-aware geographic routing in wireless sensor networks with anchor nodes
Int. J. Commun. Syst.
Unified energy-efficient routing for multi-hop wireless networks
Path loss in an Urban peer-to-peer channel for six public-safety frequency bands
IEEE Wireless Commun. Lett.
An energy-efficient clustering solution for wireless sensor networks
IEEE Trans. Wireless Commun.
Reinforcement Learning: An Introduction
Maximum lifetime routing in wireless sensor networks
IEEE/ACM Trans. Netw.
Cluster based wireless sensor network routing using artificial bee colony algorithm
Wireless Netw.
Cited by (40)
CMML: Combined metaheuristic-machine learning for adaptable routing in clustered wireless sensor networks
2022, Applied Soft ComputingImpairments-aware time slot allocation model for energy-constrained multi-hop clustered IoT nodes considering TDMA and DSSS MAC protocols
2022, Journal of Industrial Information IntegrationA Systematic Review of Quality of Service in Wireless Sensor Networks using Machine Learning: Recent Trend and Future Vision
2021, Journal of Network and Computer ApplicationsRouting in wireless sensor networks using machine learning techniques: Challenges and opportunities
2021, Measurement: Journal of the International Measurement ConfederationCitation Excerpt :It is the designer's responsibility to understand both the network structure and application and select appropriate ML techniques. Hybrid ML Techniques: Although many ML techniques are addressed in this paper to solve the routing issues in WSNs [1–146], many ML techniques can be combined and applied to solve the routing issues in WSNs, that yet to be explored. Any new routing protocol can be experimented, and simulated through any type of open-source network simulators like ns-2, ns-3 simulator, and Java-based simulator.
Fuzzy multi-hop clustering protocol: Selection fuzzy input parameters and rule tuning for WSNs
2021, Applied Soft ComputingHQCA-WSN: High-quality clustering algorithm and optimal cluster head selection using fuzzy logic in wireless sensor networks
2020, Fuzzy Sets and SystemsCitation Excerpt :Fig. 1 illustrates the single-hop and multi-hop methods. Some major applications of WSNs include monitoring forest fires, controlling the status of sensitive patients, and controlling military areas and traffic [9–12]. In general, the key factors that affect the design of WSNs are fault tolerance, scalability, costs, hardware limitations, reliability, the WSN topology, the transmission environment, and energy consumption [13–15].
Soo-Hoon Moon received the BS and MS degrees in computer science from Yonsei University, Korea, in 2008 and 2010, respectively. Since 2010, he has been pursuing his Ph.D degree in the Department of Computer Science, Yonsei University, Seoul, Korea. His current research interests include sensor networks, green networking and embedded systems.
Seung-Jae Han received the B.S. and M.S. degrees in computer engineering from Seoul National University, Seoul, Korea, and the Ph.D. degree in Computer Science and Engineering from the University of Michigan, Ann Arbor. He was a Member of Technical Staff of the Wireless Research Laboratory, Bell Laboratories, Alcatel-Lucent, Murray Hill, NJ. Currently, he is a professor in the computer science department of Yonsei University, Seoul, Korea. His research interests include mobile networking, wireless Internet, and network management.