Research On Efficient Data Gathering Algorithms For Battery-free Wireless Sensor Networks

Wireless sensors can sense data like temperature,humidity and photos from the physical world,then transmit these data to servers(or sinks)by one hop or multiple hops transmission.Wireless Sensor Network(WSN for short)is the fundamental part of the Internet of things,its applications include real-time monitoring,abnormal detection,target tracking and so on.In a traditional WSN,sensors are powered by batteries,but the amount of energy in a battery is extremely limited and we need to replace the batteries artificially with high frequency.However,in real WSN applications,sensors are deployed in hostile environments such as forests,the inner part of buildings and pine systems,in such cases,it is hard or even impractical to replace batteries frequently.Therefore,the researchers proposed to install energy harvesting devices for sensors,such devices include solar cells and Radio Frequency receivers.By doing so,sensors can harvest energy from environmental sources like solar power and radio frequency transmitters.In this scenario,sensors achieve energy self-sufficiency and operate perpetually.We term such WSN as Battery-Free WSN(BF-WSN in short).Efficient data gathering algorithms are always the focus of the academic community.There are many algorithms have been proposed for Battery-Power WSNs.But,these algorithms are ineffective in BF-WSNs.The energy support of a BF-node is unstable and uncontrollable.Thus,unlike conventional battery-power node(BP-node for short),a BF-node is not always available for data transmissions due to the energy shortage.Thus,this paper studies the efficient data gathering problem in BF-WSN.Specifically,when the computation tasks of sink are not given,we consider the minimum latency data collection problem and minimum Age-of-Information problem.On the other hand,when the computations tasks are given,we combine the transmission with computation to improve the efficiency.In this scenario,we investigate the data aggregation problem and progressive computing method.The main contributions are as follows:1.This paper studies Minimum Latency Data Collection problem in BF-WSNs.In many WSN applications,sensors need to transmit their sensor data to the sink timely and the latency minimization data collection problem is an important one in WSNs.There are some algorithms has been proposed for BF-WSNs.But no one of these existing works uses the in-network compression method,which makes the reduction on latency is slight.It has been shown that sensor data can be compressed effectively by using a linear-time compression algorithm called compressive sensing.The amount of data that needs to be transmitted can be reduced remarkably,which leads to a considerable decrease in latency.Also,the sink can reconstruct the original data series with high precision.But there is no paper studying the compressive sensing based(CS-based for short)data collection algorithm for BF-WSNs.Therefore,we proposed a distributed and energy-adaptive CS-based data collection algorithm for BF-WSN called HCS-CDS.We analyzed the performance of HCS-CDS theoretically.We simulated the solar-powered WSN and Radio Frequency powered WSN.The results show that HCS-CDS outperforms the baselines in all cases.2.We study the Age-of-Information minimization data collection problem in BFWSNs.Age-of-Information(Ao I for short)is a metric of information freshness.Recently,some works are investing in the Ao I minimization problem in BP-WSNs and single-hop BF-WSNs.But to the best of my knowledge,there is no existing work considering the Ao I minimization scheduling problem in multi-hop BF-WSNs.Therefore,we study the problem in this paper.We proved that the peak Ao I minimization problem is NP-Hard.Firstly,we focus on line topology and proposed a distributed algorithm called MAo IL.Then we consider the general topology and proposed another distributed algorithm called MAo IG.The theoretical analysis and numerical results validate that MAo IG outperforms all of the baseline schemes in all scenarios and that the experimental results tightly track the theoretical upper bound optimal solutions while the lower bound tightness decreases with the number of nodes.3.We study the minimal latency aggregation scheduling with coverage requirement problem in BF-WSNs.In some WSN applications,users are only interested in statistics of the sensor data,such as the average value of the temperature data and the maximum value of the acceleration measurements.In such scenarios,to improve the efficiency of data collection,each sensor in the network takes its sensor reading and the sensor readings it received as input and calculate the statistic locally before it sends the packets out.Such a process is termed data aggregation in WSN.The Minimum Latency Aggregation Scheduling(MLAS for short)is an important problem in WSN,we proposed to schedule a subset of the nodes in each data aggregation cycle and guarantees that the selected nodes satisfy the coverage requirement given by the user,the objective is to minimize the latency,and the problem is termed as a q-MLAS problem.We show that q-MLAS is NP-Hard and proposed an energy-adaptive and distributed algorithm.The theoretical analysis and simulation results show that our algorithm can reduce the latency effectively.4.This paper investigates the Task-Oriented Data Collection Problem in BF-WSNs.With the explosive of sensory data,how to fetch data efficiently to compute the results of tasks released by users,becomes a big issue,especially in BP-WSNs.In this scenario,the computation tasks can even be stochastic and unpredictable,which makes it very hard to know in advance what data is needed in their calculation process.Fetching all of the data that might be involved can lead to huge waste in energy and transmission resources.Therefore,we investigated the TaskOriented Data Collection Problem,that is,how to compute the accurate results of these tasks with minimum cost.In this paper,we use the downlink communication in Lora,Sigfox and NB-Io T which enable servers to guide the data collection process in sensors or divides.Specifically,we proposed a novel progressive computing method.Different from conventional methods where data fetching and computing are two separate phases,in our method,data fetching is interwoven with task computing,which enables targeted approaching to the data needed.Two example progressive algorithms for recognition and in-network query are proposed,and their performances are analyzed theoretically.Simulation results show that our algorithms can reduce the cost remarkably without any loss of accuracy.