Research On Detection And Receiving Technologies Of Communication Signals For Distributed Network

In the field of wireless communication, as we all know, there are complicated channel distortions and severe interferences and noise that degrade the qualities of received signals and eventually become the bottle neck restraints to the overall performances of the system. Communication signal processing, being developing in the last several decades, has gained a series of important achievements in coping with this category of problems and made remarkable performance improvements for current communication systems. With the prerequisite of single receiving node of the system, however, the performance improvements are limited after all. Just under this background a new technology of distributed processing based on networked multi-agent system has emerged and developing fast. Distributed processing of networked multi-agent system, due to the acquired diversity gains, can further bring about noticeable performance improvements, meanwhile needs to face and solve a series of new problems with different characteristics, which is just the theme this thesis focus on.This dissertation is devoted to a study of crucial technologies of the detection and receiving of communication signals in the environments of distributed processing for networked multi-agent system, while laying emphasis on the distributed wideband multiband signal detection, distributed modulation classification and cooperative blind equalization based on distributed particle filtering. The work finished in this thesis is a sub-project of an army engineering project undertaken by the army key lab the author works with. The main work and innovative achievements obtained in this dissertation are summarized as follows.1. The distributed algorithms for cooperative consensus optimization and adaptive estimation in distributed network, along with two related approaches independent of the network topology, are analyzed and discussed in this paper. For cooperative optimization among multipe agents, the global objective function consists of a collection of local objective functions. To achieve distributed optimization, the alternating direction method of multipliers(ADMM) is used for distributed processing after consensus constraints have been established between adjacent nodes in the algorithm based on consensus optimization. Applications of the algorithm are introduced by examples, and their performances as well as the influencing factors of the performances are analyzed by simulations. Distributed estimation algorithms based on diffusion adaptation for distributed adaptive estimation with unknown common parameters are discussed, which achieved the global cooperation estimation performance by adaptive estimation of each nodes and the information sharing between adjacent neighbors. The performance estimation of the algorithm has been provided and analyzed.2. In regard to wideband multiband signals detection, the methods based on distributed jointly sparse optimization are discussed in depth. Under thetheoretical framework of Compressed Sensing(CS), the sub-Nyquist sampling method is used to implement direct compress sampling of the wideband analog signals. Then the joint sparsity of the signals at the receiving nodes is used to improve the quality of spectrum reconstructed by cooperative processing. Targeting at the existing problems of slow convergence, heavy traffic of communication and uncertainty of convergence, a novel algorithm of distributed jointly sparse optimization is proposed, in which a mathematic model of jointly sparse optimization with consensus constraints is established and distributed processing is carried out by gradient factorization and ADMM. Simulation results show that, compared to the existing methods, the proposed algorithm has faster convergence rate and lower mean square error. Based on above work, a new detection algorithm for wideband multiband signals based on distributed jointly sparse optimization is proposed, in which the vector of mean energy in frequency domain is employed as consensus constraint and the energies of sub-channels are used for detection. Simulation results show that the proposed detection algorithm has effectively achieved spatial diversity gains and improved the overall performances markedly.3. Targeting at the problems of high computational complexity and large amount of data of spectrum reconstruction in wideband multiband signal detection, the detection algorithms are studied from the perspective of the non-reconstruction. The non-reconstruction algorithms based on subspace are studied first. Then a distributed projection approximation subspace tracking(DPAST) algorithm is proposed for effective cooperation between different nodes in distributed network. Simulation results indicate that the DPAST can effectively estimate the global signal subspace in distributed way by the use of all the signals at different nodes. Furthermore, a new detection algorithm for wideband multiband signals based on DPAST is also proposed, in which the orthogonality of the signal subspace and noise subspace are exploited to complete the blind detection. Theoretical analysis and simulation results show that the method does not need to reconstruct the original signal spectrum, thus greatly reducing the data amount and processing complexity. The fully distributed cooperative scheme is of better detection performances, network robustness and flexibility.4. As regards the modulation classification with fading channel and low SNR, the algorithms of distributed modulation classification based on signal fusion of multiple receivers are studied, in which all the received baseband signals are sent to a fusion center after synchronization of receiving nodes, and the global classification decision is made by using the hybrid maximum likelihood(HML) criterion. As a result of the spatial diversity, the performance of modulation classification are improved. Towards reducing the influence of estimation accuracy of the unknown parameter over the performance of classification and computational complexity in the HML algorithm, a joint modulation classification and parameter estimation algorithm based on Expectation Maximization(EM) algorithm is proposed. The proposed algorithm can accomplish both classification of BPSK, QPSK, 8PSK and 16 QAM signals and the estimation of unknown parameters. Simulation results show that the unknown parameters estimation based on EM estimation provides higher precision over the conventional algorithms. When the number of receivers is four and SNR of signal is more than-2dB, the average probability of correct classification is more than 95%. Then the optimal decision fusion based on maximum posterior criterion in centralized way is studied. A distributed optimal decision fusion algorithm based on belief consensus is proposed for applications distributed network. Simulation results have proved the feasibility and validity of the algorithm.5. With respect to equalization, a distributed particle filtering algorithm based on consensus optimization for cooperative blind equalization is proposed to improve the performance of bit error rate. In the proposed method, the particles are sampled according to the optimal importance function. In order to assure that all nodes have the same set of particles and weights, a conclusion is derived that the global likelihood function is the product of the local likelihood functions, based on which the consensus optimization based on alternating-direction method of multipliers and the maximum consensus can now been used to evaluate the global likelihood function. Simulation results show that only a few consensus iterations are already enough for the proposed algorithm to approach almost the same performances of their counterparts working in centralized way. Compared to the blind equalization with non-cooperation, the proposed algorithm can achieve spatial diversity gain and reduce the bit error rate. In order to further reduce the complexity of the distributed particle filtering algorithm for cooperative blind equalization, a low complexity algorithm is proposed based on minimum consensus. The algorithm, taking the mean value as an approximation of the real value of the channel, avoided the calculation of the posterior probability distribution of channel, thus greatly reduced the complexity of algorithm. Besides, the importance sampling process could be simplified greatly by taking the prior probability density as the importance function. Then the minimum consensus algorithm is used to approximate the global likelihood function for reducing the communication traffic between nodes. Simulation results show that the proposed algorithm also has achieved spatial diversity gain while the computational complexity is lower.