A new approach to grid adaption of AGIMM algorithm
Qiao Xiang-dong, Wang Bao-shu · 2003
Grid adaption is one of several approaches for model-set adaption in variable structure multiple model(VSMM) algorithm. Compared with other approaches, it is unique one with ability to produce new model-set that are not predefined and model parameter can be any value within continuous mode space. In order to improve estimation performance of system mode transition stage and steady-state stage, a new approach to grid adaption is presented in this paper. In contrast to existed grid adaption methods, grid adaption is achieved through using both model posterior probability of last scan and model prediction probability of current scan and thus information of posterior expected-mode of previous scan and predicted expectedmode of current scan are both utilized. Performance of proposed method is evaluated via two simulated example of maneuvering target tracking problem. Simulation results indicate that performance of the AGIMM algorithm, in which our new grid adaption approach is adopted, actually outperform those of algorithms in which only one kind probability is applied.