Fast Information Propagation And Active Information Acquisition In Wireless Sensor Networks

The past few years have witnessed an enormous development in 5G communications,big data and Internet of things(IoT),which brings the world into an era of information explosion.As an important component of IoT system,wireless sensor network(WSN)has to cope with the ever-rising number of sensor nodes as well as the huge amount of the generated data.Intuitively,the key to these issues lies in increasing the rate and efficiency of in-network information flows.For this purpose,it is needed to comprehensively explore the pattern of information flows and design new information transfer methodologies.In this dissertation,we mainly focus on two vital information transfer techniques in WSN,i.e.,information propagation and information acquisition.The former indicates the information dissemination process from one single sensor to multiple sensors.A new performance metric,namely,information propagation speed(IPS),is of major interest.In big data context,it is of great importance to modify the current information propagation schemes and improve the IPS given the highly dynamic network topology.As for information acquisition,it refers to the information collection process from multiple sensors to the sink node.In the face of massive information,the traditional approach which requires the collection of all data samples,is obviously inefficient and inappropriate.Therefore,in order to.reduce data redundancy and energy consumption,it is necessary to develop a more efficient and intelligent active information acquisition strategy.Motivated by the requirements above,this dissertation put forwards a fast information propagation scheme and an active information acquisition strategy.Our contributions are listed below:1.A fast virtual-MIMO-based information propagation schemeBased upon virtual multiple-input multiple-output(MIMO)techniques,we innovatively propose two fast information propagation approaches,namely cluster-based cooperative propagation scheme and greedy-combining-based propagation scheme.For each scheme,we evaluate the information propagation speed under an one-dimensional network setting.With the help of mathematical modeling and abstraction,we derive the closed-form results induced by the proposed scheme,and then analyze the impact of network density,nodes' speed and transmission range on IPS.Simulation results indicate that the two fast information propagation schemes give rise to significant performance gain compared to conventional schemes.2.An active meta-learning-based information acquisition strategyConsidering a field estimation and reconstruction problem in a wireless sensor network,we build up a communication-computation integrated framework.Furthermore,the original optimal scheduling problem is converted into a meta-learning problem of optimizing the algorithm itself,which inspires us to establish a double-layer learning machine built upon reinforcement learning as well as meta-learning.We obtain an adaptive information acquisition algorithm that can actively determine the most informative sampling location during the online field estimating process.Simulation results shows that the proposed algorithm brings a remarkable improvement in acquisition efficiency compared to conventional ones,in the meantime can be adapted to both information dynamics and task changes.