STAC: a spatio-temporal approximate method in data collection applications
Introduction
In recent years, monitoring applications have played an important role in many aspects, including target tracking [1], [2], [3], health care [4], [5], [6], [7], security surveillance [8], environmental monitoring [9], [10], [11]. Monitoring applications are often realized by WSNs or IoT. Typically, a wireless sensor network consists of hundreds or thousands of sensor nodes and many base stations (BSs) [12]. Sensor nodes report to the BS when an event is detected [13]. Nowadays, IoT utilizes WSN technologies as a necessary platform for sensing and communication of the data [14]. WSNs continue to play an important role as one of the key enabling technologies since IoT paradigm inception [15].
However, WSNs are energy constrained networks, hence, various aspects have to be considered to transmit data from each node to the destination (BS) [16]. Generally, the larger the amount of packet transmission in the network, the shorter the network lifetime [17]. Since the monitored physical world always continuously varies and WSNs nodes are often deployed redundantly to get complete monitoring data, the sensing data often have high spatial or temporal correlation [18], [19]. At present, many methods have been proposed to extend the lifetime of the network by using the spatial or temporal correlation among nodes.
In spatial, selecting partial network for data collection is one of the common methods to prolong the life cycle of WSNs [17]. To avoid coverage redundancy, some methods periodically waking or sleeping parts of nodes, so as to save energy of nodes. However, it does not consider the temporal correlation of the single node, so redundant data still exists. Other methods use spatial correlation to find the relationship among sampling data of nodes, select work nodes and use the relationship to predict the sensing data of sleep nodes. However, when the spatial correlation of nodes changes, the relationship among nodes also changes, which affects the accuracy of the predicted data. We call this phenomenon correlation-variation.
In temporal, data prediction is also one of the ways to obtain data reduction in WSNs [9]. The focus of data prediction is to use predicted data instead of real data in continuous monitoring process, to reduce the transmission of sensor nodes [9]. Therefore, data prediction can reduce the redundant data collection of each node. But, it is difficult to determine the optimal step size of data prediction.
To demonstrate our method, we take WSNs used in agriculture scenario as an example. Due to the vast area of farmland, the emergence of monitoring applications reduces the overhead of manual operation. Temperature is one of the main factors affecting the growth of crops [20], [21]. Because temperature is related to plant transpiration which affects crops growth, it should be continuously monitored to determine the right watering time. To infer the transpiration of crops, accurate and complete temperature data in the network is needed. Furthermore, several consecutive samples for single node changes smoothly and its sensing data changes periodically in a certain period of time. For instance, the temperature in a day is always the highest at noon and the lowest at night. We define a period of time corresponding to the periodic change of nodes sensing data repeatability as a work cycle. Sensors monitoring the environment around crops are located in different geographical locations, but their sensing data is similar if they are close. In addition, the position of the shadow generated by the occlusion of the object (e.g., hills and trees) changes follow the laws of nature. If the highly correlated nodes cannot be covered by shadows at a time, the correlation-variation occurs among these nodes.
For this kind of scenario and faced with above problems, this paper proposes a new energy-saving and efficient method named spatio-temporal approximate data collection (STAC). STAC reduces the sampling and transmission of redundant data to extend the network lifetime. And it restores data that has not been sampled while meeting the accuracy requirements, ensuring the integrity of the data.
Firstly, this paper minimizes the number of work nodes in each work cycle by sleeping some nodes. This reduces the amount of data collection and transmission, and achieves a balanced energy distribution in the network. By finding out the relationship among nodes, the prediction function of a node is formed. This prediction function is called the recovery function of this node, and we use it to predict the data of this node if it sleeps. Faced with the situation of correlation-variation, this paper proposes a verification mechanism to ensure the accuracy of predicted data. Secondly, in order to further reduce the data redundancy of nodes, we dynamically adjust the sampling interval of nodes through Q-learning [22]. In Q-learning stage we consider the “change rate of sampling data” and the remaining energy of nodes to make the setting of sampling interval more reasonable. At the end of a work cycle, STAC predicts the missing data which is not collected due to the dynamic change of sampling interval. STAC ensures that the error between predicted data and real data is tolerable. Besides that, our method avoids the premature death of the single node, which is conducive to prolonging the life cycle of the network.
The main contributions of this paper are summarized as follows:
A spatio-temporal approximate data collection method is proposed. By minimizing the number of work nodes in each work cycle and adjusting the sampling frequency of nodes, this method maximizes network lifetime and balances energy distribution.
We find and define the correlation-variation. A verification mechanism for correlation-variation is proposed, which further improves the accuracy of the predicted data.
A large number of simulations have been carried out under different conditions. Through the verification and analysis of the simulation results, we prove the superiority of STAC.
The reminder of this paper is organized as follows. In Section 2, we survey the related literature. The details of STAC and some important definitions are given in Section 3. In Section 4, we discuss and analyze the simulation results. Finally, Section 5 reports the concluding remarks.
Section snippets
Related work
This section introduces the literature related to the method of prolonging the lifetime of WSNs. The existent research regarding the correlation in the network can be classified into the following three categories.
Problem definition and method description
This section is divided into five parts. In the first part, we give an overview of STAC and introduce some important definitions in this paper. The second part introduces the formation of recovery function and the prediction of missing data. The third part describes the determination of the correlation-variation and the verification mechanism of it. Then, the fourth part introduces the selection strategy of work nodes. The dynamic adjustment strategy of sampling interval is given in the last
Experiment and performance evaluation
This section presents the numerical results of the performance of the proposed STAC. We first describe the simulation parameters and two baselines. Then we demonstrate the effect of our spatio-temporal approximate data collection method on lifetime and loss of accuracy with real-world data set. In this way, we utilize MATLAB for evaluating the performance of STAC. We compare the work of this paper with OPR and OIQL mentioned in Sections 2.3 Methods using spatio-temporal correlation, 3.5
Conclusions
In this paper, we proposed a new framework for long-term monitoring applications, which depends on WSNs or IoT and utilizes data prediction, selecting work nodes, and Q-learning. In our framework, we considered the spatio-temporal correlation among nodes. Firstly, we combined data prediction with selecting work nodes in each work cycle. Secondly, Q-learning was used to adjust the sampling interval of nodes. Finally, the BS predicted the whole data with error tolerance. Simulation results by
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is supported by the National Natural Science Foundation of China (NSFC) (Grants No. U19A2061, No. 61772228, No. 61902143), National key research and development program of China (Grant No. 2017YFC1502306), Scientific and Technological Development Program (Grant No. 2020122208JC), Research Project by the Education Department of Jilin Province (Grant No. JJKH20211105KJ).
References (36)
- et al.
Multiple-target tracking based on compressed sensing in the Internet of Things
J. Netw. Comput. Appl.
(2018) - et al.
Optimizing K-coverage of mobile WSNs
Expert Syst. Appl.
(2018) - et al.
CUIDATS: An RFID-WSN hybrid monitoring system for smart health care environments
Future Gener. Comput. Syst.
(2018) - et al.
A lightweight and robust two-factor authentication scheme for personalized healthcare systems using wireless medical sensor networks
Future Gener. Comput. Syst.
(2018) - et al.
A body sensor data fusion and deep recurrent neural network-based behavior recognition approach for robust healthcare
Inf. Fusion
(2020) - et al.
Data prediction, compression, and recovery in clustered wireless sensor networks for environmental monitoring applications
Inform. Sci.
(2016) - et al.
Wireless sensor networks for agriculture: The state-of-the-art in practice and future challenges
Comput. Electron. Agric.
(2015) - et al.
A sleep scheduling approach based on learning automata for WSN partialcoverage
J. Netw. Comput. Appl.
(2017) - et al.
Maximizing the wireless sensor networks lifetime through energy efficient connected coverage
Ad Hoc Netw.
(2017) - et al.
Prediction-based data aggregation in wireless sensor networks: Combining grey model and Kalman Filter
Comput. Commun.
(2011)