Research On Distributed Collaborative Estimation Method Based On Bias-compensated LMS Algorithm

In recent years, the distributed cooperative information processing technology is widely used in the field of wireless communication networks, such as wireless sensor network target location, multi UAV target search, multiple spacecrafts formation control and industrial wireless network parameter estimation, etc. Distributed cooperative information processing means that the information of the whole network can be processed through the different information interaction models between agents. In this paper, we mainly study the application of distributed cooperative information processing technology in adaptive target parameter estimation, that is, for the unknown parameters, and based on the adaptive filtering algorithm, we use distributed strategy to estimate the target parameters.Due to the LMS adaptive filtering algorithm has the advantages of small computation and simple structure, the whole study is based on the LMS algorithm, and we achieve the performance comparison of the algorithms based on FIR and IIR two adaptive filter. Firstly,we introduce the adaptive estimation of single agent, and the LMS adaptive filtering algorithm is theoretically analyzed. On the basis of single agent adaptive estimation, this paper studies the multi-agents distributed cooperative estimation algorithm, and introduces the multi-agent network model and the distributed cooperative strategy in detail, which includes three kinds of incremental strategies, the consistent strategy and the diffusion strategy. we analyze the information of different strategies, and give the distributed LMS algorithm based on LMS algorithm, and the accuracy of the algorithm is compared with by simulations. Simulation results show that compared with the single agent non-cooperative estimation algorithm, the distributed cooperative estimation algorithm has better estimation accuracy, and it also shows that the performance of the three kinds of distributed cooperative estimation algorithms are different under different step-size conditions.In consideration of the actual environment, the input and output signals of the distributed network usually contain many kinds of noise(such as quantization noise and measurement noise). These results have some serious errors. Therefore, the relevant research is based on the principle of bias compensation, through the elimination of the noise caused by the deviation, a distributed cooperative estimation error compensation algorithm is proposed, which can obtain unbiased estimates. While in previous studies, it isusually assumed that the noise variance is known or contains constraints, and can not achieve real-time estimation of the noise variance without constraints. In this paper, based on previous studies, a new method of real-time estimation of noise variance is proposed.Based on different distributed LMS algorithm, a distributed bias-compensated LMS algorithm is proposed.The simulation results show that, compared with the distributed LMS algorithm,distributed bias-compensated LMS algorithm can achieve unbiased estimation of target parameters, and then compared with the previous bias-compensated LMS algorithm, the bias-compensated LMS algorithm proposed in this paper has better estimation accuracy and lower mean square error(MSD), and the results can be achieved without bias estimation of different SNR levels.