Knowledge Aided Sieving Multiple Model Maneuvering Target Tracking Algorithm

The technology of target tracking has received a broad application in numerous domains.For instance,in monitoring and piloting the general aviation.It can be used to obtain: target’s position,speed,motion trajectory,and others.Real-time and effective tracking of maneuvering targets has always been a research hot spot.It’s also a challenging topic in the academic world,and applied branches.The multiple-mode algorithm is the most commonly used and performant traditional maneuvering target tracking method.However,the multiplemodel algorithm generates mismatch problems.Examples of such problems are: measurement noise,measurement model,motion model set,transition probability matrix(TPM),and model probability(MP).The problems arise under conditions of unidentified measurement noise model,meteorological environmental impact,uncertainty of target classification,terrain environment,and obstruction of obstacles.In addition,there’s also mismatches between real values,and others.This in so doing will damage and diminish the tracking performance of the multiple-mode algorithms.The difficulties and troubles caused by the problems mentioned above can be eluded via the knowledge-aided multiple-mode algorithm.It can effectively develop the tracking performance.However,other problems equally exist.For instance,how to introduce frequently used external knowledge information,like: target classification,obstacle information,measurement accuracy range,and others.How the external knowledge information affects: the motion model set,measurement model,transition probability matrix,model probability,etc.of the multiple-model algorithm.For these problems to be solved,this thesis studies both multiple-mode algorithm and knowledge-aided sieving combined and used in maneuvering target tracking.This thesis proposes knowledge-aided multiple-mode maneuvering target tracking algorithms.This research study improves and develops knowledge-aided multiple-mode algorithms.The four following points,reflect the key contributions and innovations:(1)The principle behind the knowledge-aided sieving multiple-model algorithm is studied,alongside its proper use,capable of improving the performance of the multiple-model algorithm.The elementary principle of the multiple-model algorithm,the iterative processes,classifications,and others,of the commonly used third-generation multiple-mode algorithms,are studied.The present knowledge-aided multiple-model algorithm is classified and introduced with respect to different functional elements.Its advantages,disadvantages,and applicability are compared and analyzed.(2)The study of the utilization of knowledge-aided sieving in the selection of motion model sets,is done.Two algorithms are proposed,comprising using the target classification and the obstacle information to adjust the motion model set.The diversity of general aviation aircraft and the large variation in maneuverability renders the model set in the multiple-model algorithm too bulky.This leads to a drop-in performance.To solve this problem,this thesis proposes a Target Classification Aided Multiple-model(TCA-MM)algorithm,using target’s classification information to get model set.The algorithm uses the acquired target classification information to sieve through the model set.This is to diminish the error model and the competition between the models.The results reveal that,the algorithm can enhance the tracking accuracy,and diminish the computational load.Since AGV is limited by obstacles in an internal part,causing difficulties in tracking.This thesis proposes a Model Set Adaptive Multiple-model(MSA-MM)algorithm.After abridging the hindrances into polygons,the algorithm screens the model set by using the field angle of the target’s projected position to the obstacle.This is to attain an adaptable modification of the model set,in order to enhance the accuracy of multiplemodel algorithm tracking.This hence,reduces the number of errors.The results illustrate that,in an internal environment with additional obstacles,the algorithm can accomplish a better tracking effect.(3)The use of knowledge-aided sieving in the selection of measurement models.Both a variation Bayesian multiple-model algorithm,and a knowledge-aided AGV multiple-model algorithm are proposed.In general aviation aircraft tracking,the measurement noise of ADS-B data is uncertain,the maneuverability of these targets is solid.Therefore,it’s difficult to track.Aiming to solve this problem,this thesis proposes a variation Bayesian interactive multiplemodel(VB-IMM)algorithm,using the variation Bayesian method to estimate the measurement noise variance.This equally uses the Navigational Accuracy Category for position(NACp)aided in ADS-B data,and then performs multiple-model filtering.The results show that the algorithm can effectively estimate the unknown measurement noise and obtain better tracking results.When the AGV,which uses ultra-wideband positioning,uses the Chan positioning model with TDOA measurement,there is ambiguity and uncertainty in the positioning near the base station.To solve this problem,this thesis proposes a knowledge-aided AGV multiplemodel(KA-AGVMM)algorithm,using target motion information and environmental information.When the AGV employs three positioning base stations for tracking,the algorithm uses motion information and environmental information to formulate three constraint rules to exclude the error tracking results.The results show that the tracking algorithm can effectively eliminate wrong tracking and improve the tracking accuracy.(4)The application of knowledge-aided in the adjustment of the TPM and the MP.For the problems where general aviation aircraft are restricted by buildings,when flying at low altitude,leading to difficulties in tracking and many incorrect estimates.This thesis proposes the use of obstacle information fuzzy inference,to adjust MP multiple-model(OATPM-MM)algorithm,MP multiple-model(OAMP-MM)algorithm and to simultaneous adjust the multiple-model(OABoth-MM)algorithm.The OAMP-MM algorithm uses the distance and field angle of the predicted position to the obstacle areas,the weight of the model is acquired by fuzzy inference to adjust the MP.The OATPM-MM algorithm uses fuzzy inference to obtain the expected residence time of the model to adjust the TPM.The simulation experiment and the measured experiment indicate that,these algorithms can improve the accuracy of model discrimination,reduce the number of wrong estimates,and improve the tracking accuracy,compared to other related algorithms.The series of knowledge-aided multiple model sieving algorithms are proposed in this thesis.They have been studied on the theoretical framework.Their validity and superiority of the proposed algorithm are verified by the corresponding simulation and measured experiments.Starting from the practical application,the mismatch problems between measurement model,motion model set,transition probability matrix,model probability,etc.and the real value,are solved.This thereby improves the tracking performance of the target tracking.