Link Sampling Technology Based On Compressive Sensing And Modulating Retro-reflector

The sensing data in wireless sensor network are just the observed values of several random positions in monitoring region.According to these discrete observations,user needs to reconstruct the gridding environment information of monitoring region.The commonly method of discrete data gridding is spatial interpolation.Under the framework of compressive sensing,the premise of environment information can be reconstructed accurately includes that the signal is sparsity,the positions of nodes are randomness and the data of nodes is redundancy.Compressive sensing provides a technical approach for discrete data grid.The purpose of this dissertation is mainly for two optimization problems: 1)for the wireless sensor networks with large scale and sparse nodes,how to reconstruct all the environmental information of the monitoring area by using the sparse sampling data;2)for the large scale dense sensor networks,how to transmit the least sensor node data without losing the environment information.Focusing on the two issues above,the main work and innovations of this dissertation include:For the first issue,it proposes a scheme of discrete data gridding: scheme of signal reconstruction with information correction layer-by-layer for sparse sampling.It fills the blank grid of monitoring region based on compressed sensing theory.The advantage of the method is that it requires fewer sampling points than the interpolation based discretization method.This scheme assumes that sensor network is equivalent to the sparse sampling of physical environment,so it can use compressed sensing algorithm to realize discrete data gridding.The scheme includes four sub-algorithms as following:1)A new method of measurement matrix construction based on Empirical Mode Decomposition of reference signal is proposed.It uses the Intrinsic Mode Function to construct a cyclic matrix.Due to the Intrinsic Mode Function circulant matrix can reflect the structure characteristics of the signal,it is very suitable for the reconstruction of sparse sampling.This dissertation prove that this matrix satisfies the restricted isometry property condition by Ger?gorin disc theorem.Using denoising effect as the estimating standard,it simulates the denoising process of ideal signal and actual signal.The simulation resultsshow that:(a)adding a certain degree noise to reference signal,it has a better effect on reducing noise;(b)while the noisy signal and the reference signal has a dislocation in the time domain,compared with the ideal condition,the denoising effect is significantly lower,but the frequency features are still highlight.2)Iteration forecast orthogonal matching pursuit algorithm is proposed.It is based on the relationship between the two atoms of two adjacent iterations,and their inner products with the earlier residual have interval feature,which can support the prediction of the atom in the next iteration with high probability.Experimental results show that,compared with other matching pursuit algorithm,its time-consuming is greatly reduced while the reconstruction accuracy has not significantly reduced.3)Joint optimal orthogonal matching pursuit algorithm is proposed,whose innovative feature is its higher reconstruction accuracy exchanged by waste of time,and it is suitable for the situations that need higher percentage of input signals recovered correctly with less number of measurements.4)Reconstruction algorithm with information correction layer-by-layer is proposed,which is based on the Intrinsic Mode Function circulant matrix and improved orthogonal matching pursuit algorithm.The algorithm defines the layer as: using the estimated signal obtained from the upper layer to build the Intrinsic Mode Function circulant matrix,the signal reconstruction is carried out after observing the sensor data,then using sensor data to replace the corresponding value of reconstructed signal to generate a estimated signal of new layer.The algorithm can get grid data with higher correlation degree with environment information,and needs less sampling points than interpolation and traditional compressed sensing.For the second issue,it proposes a new data collection scheme: link sampling.Different from the existing data fusion or removing redundant collection methods to reduce the amount of data transmitted,under the framework of compressive sensing,link sampling is a data collecting scheme by the random establishment of communication link between the sink and sensor nodes,and reconstruct the gridding signal of monitoring region directly.The essence of link sampling is to carry out the random sparse sampling of the nodes in the large scale wireless sensor networks.This scheme includes two aspects as following:1)Real time link sampling algorithm is proposed.In the data collection process,through the real-time detection of whether the gridding data can reflect the real environment information the signal,the algorithm adjusts collection strategy dynamicly to ensure the least transmission of nodes data.It has a unique advantage in extending the wireless sensor network lifetime.2)Physical implementation scheme of link sampling is proposed,which randomly establishes the communication link between the sink and sensor nodes supported by modulating retro-reflector and optical space division multiple access.Modulating retro-reflector optical communication system is actually a request response communication mode: the request side has option on the establishment of communication link.Only the response side that irradiated by laser can transmit their data to the request side by reflected laser.So in the scheme,it sets sink node as request side and sensor node as response side.The sink collects sensor data in batches by emitting(and receiving)multiple laser beams through optical spatial division multiple access antenna.By simulations and experiments,the two key technologies that modulating retro-reflector and optical space division multiple access are researched,which lay the foundation for the whole physical realization of link sampling.Real time link sampling scheme compromises all the algorithms of this paper.Its implementation enables wireless sensor networks can adaptively determine the spatial sampling ratio of the sensor network for different perceptual objects.While reducing the number of nodes participating in data transmission and opening a new way for prolonging WSN life,it can also guarantee that the signal obtained by gridding the discrete data can reflect the real environment information.