Structured Compressed Sensing Based Traffic Prediction In Wireless Communication Networks

The newly proposed concept of green communications makes the energy efficiency improvement in communication networks become a research hotspot. Unfortunately, in the current cellular networks, the power of base stations is designed according to the requirements in one cell’s peak user traffic. Consequently, the designed transmit power might be not in accordance with the traffic variation, thus leading to an energy wastage. Therefore, to accurately predict the traffic variation in cellular networks provides the foundation for the dynamic allocation of energy resources and the energy efficiency improvement. The researches before have validated that the temporal and/or spatial pattern in user behaviors make the cellular network traffic vary periodically. One of the results is that the cellular traffic data possesses the redundancy and temporal and/or spatial structural property. On the other hand, this redundancy makes the traffic matrix low-rank. In other words, we can regard a traffic matrix as a sparse matrix.Compressive sensing is one novel means to deal with large amount of data with sparse property. Its core methodology is to make the sampling and compressing of data into one step. To be specific, it concurrently samples and compresses from a signal at a rate from lower than Nyquist rate of this signal and reconstruct the signals at the destined receiver while employing the sparsity of signals. The ability to effectively sample-compress and precisely reconstruct the signal with mass data make compressive sensing widely applied in wireless communication networks.This paper firstly details the fundamentals of compressive sensing, including the mapping of a signal to its sparse bases, the design of the measure matrix and the design of algorithms to reconstruct a signal. Meanwhile, this paper also introduces the application of compressive sensing in some layers of wireless networks, including spectrum sensing and channel estimation in physical layer, the data collection and network monitoring in network layer and abnormal event detection and user location in application layer.This paper concentrates on the researches of compressive sensing based on chunk sparse signals and the application of temporal/spatial compressive sensing in traffic loads prediction of cellular networks. The realistic signal processing usually makes the signal carry a large amount of structural information, especially chunked sparse signals. Therefore, this paper proposes a methodology, including the sparse model definition of chunk-based sparse signals, the design principles and means of observation matrix and the design of reconstructing algorithms. Meanwhile, the proposed methodology is applied to traffic load prediction in cellular networks by taking advantage of temporal and/or spatial structural property in traffic matrix. In other words, a temporal and spatial compressive sensing method is proposed and is exploited to precisely predict the variation in cellular networks.