資源描述:
《免疫粒子群粒子濾波算法及硬件實(shí)現(xiàn)研究》由會(huì)員上傳分享,免費(fèi)在線(xiàn)閱讀,更多相關(guān)內(nèi)容在教育資源-天天文庫(kù)。
1、浙江大學(xué)控制科學(xué)與工程學(xué)系碩士學(xué)位論文免疫粒子群粒子濾波算法及硬件實(shí)現(xiàn)研究姓名:徐濤申請(qǐng)學(xué)位級(jí)別:碩士專(zhuān)業(yè):系統(tǒng)工程指導(dǎo)教師:馬龍華20100101摘要在非線(xiàn)性濾波領(lǐng)域,粒子濾波算法在繼傳統(tǒng)的擴(kuò)展卡爾曼濾波算法和無(wú)跡卡爾曼濾波算法之后開(kāi)始得到人們的重視,并廣泛應(yīng)用于目標(biāo)跟蹤、導(dǎo)航制導(dǎo)與控制、圖像處理及故障檢測(cè)等領(lǐng)域。粒子濾波是一種基于蒙特卡洛模擬實(shí)現(xiàn)遞推貝葉斯濾波的技術(shù),利用在狀態(tài)空間中傳遞的隨機(jī)樣本,對(duì)狀態(tài)后驗(yàn)概率密度函數(shù)進(jìn)行近似,并根據(jù)蒙特卡洛估計(jì)原理估計(jì)狀態(tài)值。本質(zhì)上適用于任何非高斯及非線(xiàn)性的情況。但是,粒子濾波也存在固有的缺陷,如粒子退化、樣本貧
2、化及計(jì)算速度慢等問(wèn)題。本論文針對(duì)傳統(tǒng)粒子濾波技術(shù)的固有缺陷,研究對(duì)粒子濾波算法的改進(jìn)及其硬件實(shí)現(xiàn)問(wèn)題,主要研究工作包括:1,針對(duì)傳統(tǒng)粒子濾波算法粒子退化及樣本貧化的固有缺陷,利用免疫粒子群優(yōu)化思想對(duì)粒子濾波的重采樣環(huán)節(jié)進(jìn)行優(yōu)化處理,提出免疫粒子群粒子濾波算法(IMPSOPF),通過(guò)狀態(tài)估計(jì)及目標(biāo)跟蹤應(yīng)用的仿真研究,證明該算法在解決粒子退化問(wèn)題的同時(shí)避免了樣本貧化現(xiàn)象,算法是有效的。2,針對(duì)粒子濾波算法計(jì)算速度慢的固有缺陷,本論文通過(guò)MATLAB輔助ISE的設(shè)計(jì)方法將串行算法進(jìn)行模塊劃分及Vetilog語(yǔ)言描述,并在賽靈思V5系列FPGA中實(shí)現(xiàn)IMPSOP
3、F算法。硬件算法的目標(biāo)跟蹤應(yīng)用的仿真研究表明,基于FPGA的IMPSOPF算法在保證算法估計(jì)狀態(tài)精度的同時(shí),極大的提高了算法的運(yùn)算速度,為今后IMPSOPF算法應(yīng)用于目標(biāo)跟蹤、導(dǎo)航制導(dǎo)與控制等實(shí)時(shí)性要求較高的領(lǐng)域奠定了堅(jiān)實(shí)的理論基礎(chǔ)。關(guān)鍵詞:粒子濾波;免疫粒子群優(yōu)化(IMPSOPF);FPGA;并行計(jì)算處理AbstractInnonlinearfiltering,particlefilteralgorithmhasbecamemoreandmorepopularafteralgorithmsofExtendedKalmanFilterandUnscent
4、edKalmanFilterwhichiswidelyusedinthefieldsofobjectivetracking,navigationguidanceandcontrol,imagineprocessingandmalfunctiondetecting.ParticlefilterisakindofmixtureofMonteCarloestimationtheoryandtechnologyofBayesfilteringwhichismakinguseofaswarmofparticlestransportedinpossibleareas
5、toestimatethefunctionofpriordensity.Inessential,itfitsforanynon—Gaussandnonlinearsituations.Butparticlefilteralgorithmhasitsowndrawbackssuchasparticledegenermion,impoverishmentandslowvelocityincalculming.Theinherencedrawbacksofparticlefilteralgorithmaretakenintoconsiderationinthi
6、spaperwithimprovedparticlefilteralgorithmandhardwareaccomplishmentaspaperresearchobjectives.Themainresearchworkisasfollowing:1,Takingparticledegenerationandsampleimpoverishmentintoaccount,proposesimmunityparticleswarmoptimizationparticlefilteralgorithm(IMPSOPF)whichcandealwithbot
7、hparticledegenerationproblemandsampleimpoverishment.Simulationinthefieldsofstateestimatingandobjectivetrackingshowitseffectiveness.2,Takingparticlefilteralgorithmslowvelocityofcalculatingintoaccount,onthebaseofmethodofMATLABaidingISEdesigning;wedecomposetheparticlefilteralgorithm
8、intosomemodulesandthenaccomplishfunction