Procjena stanja nelinearnih dinamičkih sustava s neodređenostima

doctoral thesis

• Title (english): State estimation of nonlinear dynamic systems with uncertainties
• Author: Jadranko Matuško
• Mentor: Nedjeljko Perić (mentor)
• Committee member: Ivan Petrović (član povjerenstva)
• Committee member: Nedjeljko Perić (član povjerenstva)
• Committee member: Joško Deur (član povjerenstva)
• Committee member: Branko Jeren (član povjerenstva)
• Committee member: Vesna Županović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Control and Computer Engineering); Zagreb
• Defense date and country: 2008-04-25, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering; Automation and Robotics
• Universal decimal classification (UDC): 62 - Engineering. Technology in general ; 007 - Cybernetics. Self-acting systems

Abstract

U ovom je radu razmotren problem procjene stanja dinamičkih sustava s neodređenostima s posebnim naglaskom na dvije vrste neodređenosti: modelske i stohastičke. Dan je pregled najvažnijih postupaka procjene stanja sustava s neodređenostima. Predložen je postupak procjene stanja sustava s modelskim neodređenostima zasnovan na neuronskim mrežama koji je primjenjiv na široku klasu nelinearnih sustava. Za predloženi je procjenitelj stanja dokazana stabilnost i konvergencija procjene korištenjem Lyapunovljeve analize stabilnosti. Također su napravljena dodatna strukturna pojednostavljena procjenitelja stanja za dva posebna slučaja. Kvaliteta predloženog procjenitelja stanja sustava simulacijski je provjerena na primjeru procjene sile trenja između automobilskog kotača i podloge. Nadalje, u radu je predložen i višekriterijski postupak adaptacije broja čestica čestičnog filtra kojim se znatno smanjuju numerički i memorijski zahtjevi postupka procjene. Pored adaptacije broja čestica predloženi postupak omogućuje oporavak postupka procjene u slučajevima degeneracije samog postupka. Predloženi postupak adaptacije broja čestica filtra primijenjen je na rješavanje problema lokalizacije mobilnog robota. Kroz simulacijske i eksperimentalne provjere potvrđena je primjenjivost i robusnost predloženog postupka pri globalnoj lokalizaciji i rješavanju problema tzv. "otetog" robota. Prilikom izrade ovog rada korišteni su Matlab/Simulink i Player/Stage razvojna okruženja.

Abstract (english)

In this thesis the problem of state estimation of nonlinear dynamic systems with uncertainties is considered. Special attention is given to the two particular types of uncertainties: model and stochastic uncertainties. Overview of the existing approaches to the state estimation of nonlinear systems with uncertainties is presented. A novel approach is proposed for the state estimation based on neural networks. This approach is applicable to the wide class of nonlinear systems. Stability and convergence of the proposed neural network based state estimator is proven via Lyapunov stability analysis. Additional structural simplifications of the proposed estimator are given for two special cases of nonlinear systems. The quality of the proposed state estimation procedure is tested on the problem of tire/road friction force estimation. A new multi-criterion procedure is proposed for adaptation of the particle filter sample size. This new approach leads to the significant reduction in the particle filter numerical and memory requirements. The proposed adaptation procedure is used in the problem of mobile robot localization. The robustness and applicability of the adaptation procedure for both global localization and "kidnapped" robot problems is confirmed via simulation and experimental tests. The developed algorithms were tested within Matlab/Simulink and Player/Stage programming environments.

• Keywords: nelinearni sustavi; modelska neodređenost; stohastička neodređenost; procjena stanja; neuronske mreže; čestični filtri; Monte Carlo postupci
• Keywords (english): nonlinear systems; model uncertainty; stochastic uncertainty; estimation; neural networks; particle filters; Monte Carlo methods
• Language: croatian
• URN:NBN: urn:nbn:hr:168:103881
• Study programme: Title: Electrical Engineering and Computing; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti elektrotehnike i računarstva (Doktor znanosti elektrotehnike i računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-05-30 13:55:55