Multi-Agent Systems State Estimation Based On Quantized Information

Focused on the state estimation problem of wireless sensor networks without fusion center, an algorithm of distributed Kalman filter using quantized information (QDKF) is introduced.Firstly, the states of the target are estimated with a distributed Kalman filter. A dynamic optimal method to choose the consensus weight matrix is used to weight the significance of each neighbor node’s information according to their estimation accuracy. The object of the optimization is that the whole networks share the estimation with a minimum error covariance matrix (ECM).Furthermore, considering bandwidth constraint of the network, the information is quantized before communication to reduce requirement on the network bandwidth. The expression of the new system noise and measurement noise is given and proved. Simulations are conducted by using the proposed QDKF algorithm with an eight-bit quantizer. Compared with the Metropolis weighting and the maximum degree weighting methods, the proposed dynamic weighting method reduces 25% and 27.33% of the estimation root-mean-square error, respectively.The simulation results show that the QDKF algorithm can improve the estimation accuracy and reduce the requirement of network bandwidth, and is applicable for limited communication scenario.