| With the improvement of application value, the multi-sensor system has become ahot problem of domestic and foreign research, whose application field has extendedfrom military, national security to civil fields, such as disaster prediction, health care,traffic management, urban informatization construction, public transportationmanagement and so on. In addition, Networked Control Systems(NCSS), which arecombination of obtaining information, fusion estimation, and system control, with acharacteristic of forming a closed-loop controlled system by network, are greatlyfavored in the application of underwater torpedo guidance and control, air-to-airmissiles collaborative combat, communications, industrial control, and military andcivil fields. This thesis puts forward distributed fusion unscented kalman filter anddistributed fusion particle filter for multisensor nonlinear systems with multiple packetdropouts, respectively.The design of distributed fusion unscented kalman filter, first considers observedpacket dropouts or state packet dropouts occurring at the k time will result in losingthe state estimation of k time, and causing the unscented kalman filter can not proceedthe recursive filtering at the k+1time. This thesis adopts the state of the previous timeby one step to forecast, which approximately replaces the state estimation of the currenttime, and then proposes the multisensor distributed fusion unscented kalman filter.The design of distributed fusion particle filter, due to particle filter processes thefilter by reproducing the sampling particles; the packet dropouts of state estimation atk time couldn’t affect the process of particle filter at k+1time. Therefore, accordingto the problem of observed packet dropouts occurring at the k time, this thesis adoptsto copy the weight of resampling particles in the previous time, estimates the probabilitydistribution of important particles, and then proposes the multisensor distributed fusionparticle filter. |
PDF Full Text Request