Distributed Quantized And Dimensionality Reduction Fusion Eestimation For Networked Multi-Sensor Systems With Bounded Noises

Information fusion technology can process the multi-sensor data from multiangle and multi-levels by some optimize rules to get a consistent description and explanation.The result will have a higher accuracy than single sensor.Distributed fusion estimation is an important branch in information fusion technology.A local estimate need to be firstly obtained and then be sent to fusion center through communitation network.Finally,the distributed fusion estimation will be computed in fusion center.So compared with signal processing from single sensor,distributed fusion estimation has has faster compute speed,higher reliability and easier implement,it has been widely applied in practice.However,with the increase of the number or the distribution of the sensors,network bandwith became the bottleneck of the estimate perfoemance.Besides,Kalman filter requires the statistical property to be known in advance.So,studing distributed fusion estimation under bandwidth constraint has a significant meaning.Because fusion center cannot receive the complete sensor information under the bandwidth constraint,the estimation performance will decrease.At the same time,although H?infinity method shows a well performance when the statistical property of the system noise is unknown,it requires the noise is energy bounded.As for the bounded noise which is bounded at every time but does not satisfy energy bounded,a bounded recurcive method was ultilized in this paper to handle it.By combining the quantized method and dimensionly reduction method,a distributed fusion estimation algorithm with bounded noises under bandwidth constraint was proposed.The main work and results can be summarized as follows:(1)Deeply analysing the Kalman filtering based distributed fusion estimation method,and pointing out its disadvantage.Moreover,because the statistical property is difficult to be known in advance,a bounded recursive based distributed fusion estimation was introduced.At last,a moving robot simulation is given to prove the effectiveness of the proposed methods.(2)The unstable system whose states may grow to infinity was studied.Bounded recursive based distributed fusion estimation with innovations for unstable systems was proposed.A satisfactory estimation performance can be achived by this algorithm when the system states are infinity.An unstable system example is given to demonstrate the effectiveness of the proposed methods.(3)The nonlinear dimensionality reduction distributed fusion estimate problem is investigated under the communication bandwidth constraints.By linearing the nonlinear state space equation and combining the dimensionality reducing method,the distributed fusion estimation for nonlinear systems with bounded noises was obtained.A sufficient condition that the estimator was stable would be obtained if the local estimation was stable.Dimensionality reduction is modeled by a group of random variables.An unlinear mobile robot system example is given to demonstrate the effectiveness of the proposed methods.Finally,the thesis has been summarized and future work is presented.