Research On Adaptive Algorithm For High Maneuvering Target Tracking

Target tracking is the core issue of military research. In recent years, thereappeared a variety of aircrafts with high maneuvering ability such as combat aircrafts,missiles, and naval ships. The truth shows that traditional target tracking methods nolonger totally meet the adaptability and accuracy requirements of high maneuveringtracking. At the same time, the rise of passive sensors for maneuvering tracking bringsthe problem of correlated measurement noises.Based on the background above, this paper aimed at designing target trackingmethods with better adaptability and accuracy. We clarified the definition of strongmaneuverability and analyzed single model tracking methods and multi-modeltracking methods separately to develop adaptive tracking algorithm. Furthermore, wepresented correlated noise decoupling algorithm based on eigenvalue decompositionand orthogonal projection. The specific research contents are as follows:1) Clarified the definition of strong maneuverability and summarizes threemeasures of it. Enumerated actual maneuvering trajectories such as J turning.Introduced linear and nonlinear state estimation methods, and pointed out that Kalmanfilter loses sensibility for maneuvers when it reach steady state, and introduced strongtracking filter which can fix this problem. Then presented performance evaluationindicators of target tracking algorithm.2) Studied single model maneuvering target tracking adaptive algorithm.Analyzed traditional algorithms such as Singer model, and pointed out the limit offixed-parameter model in changing environment adaption. Based on existingalgorithms, we proposed maneuvering frequency adaptive algorithm based onmaneuvering time accumulation, acceleration variance adaptive algorithm based onSage-Husa method, and system variance adaptive algorithm based on double filtering.Tangential and normal acceleration model was presented to better describe themaneuverability at last. Simulation was implemented to verify the performance.3) Studied multiple-model maneuvering tracking algorithm. Multiple-modelmethod is a better choice in target tracking while singer model cannot fit the varietyof motion. Based on the variable structure model theory and model set selectiontheory, we introduced variable structure mode algorithm. Then we proposed three variable structure methods: active digraph algorithm with tangential and normalacceleration; digraph switching algorithm with parallel structure; adaptive gridalgorithm with crossover structure. And then simulation was implemented to verifythe effect.4) Studied the multi-sensor target tracking algorithm with correlated noises.Proposed fusion algorithm and sequential algorithm based on eigenvaluedecomposition, and discussed the effect on computation reducing. Proposedsequential fusion method and simplified method based on orthogonal projection, andsimulation was implemented. At last, we presented the multi-sensor trackingalgorithm flow with single model or multiple models.