LQG/LTRå¶åŸ¡ãåŠãã
LQG/LTR (Linear Quadratic Gaussian / Loop Transfer Recovery) ãšã¯ LQGãšã¯ãã«ã«ãã³ãã£ã«ã¿ãçšããŠæšå®ããç¶æ ã«å¯ŸããŠãæé©ã¬ã®ã¥ã¬ãŒã¿ãçšããŠç¶æ ãã£ãŒãããã¯ããããªããããç¥ãããå¶åŸ¡æ³ã§ãã LQGãæéé åã§ã®å¶åŸ¡åšèšèšã§ããã®ã«å¯ŸããŠãåšæ³¢æ°é åã§ã®èšèšãèæ ®ããã®ããLQG/LTRã§ãã ä»åã¯Stein G and Micheal A. The LQG/LTR procedure for multivariable feedback control designãåèã«ããŸããã åé¡èšå® ä»ãå¶åŸ¡å¯Ÿè±¡ãäŒéé¢æ°è¡å$G(s)$ãšããŠã¢ããªã³ã°ãããŠãããšããŸãã ããã§ãå¶åŸ¡å¯Ÿè±¡ã¯éæå°äœçžç³»ã§ãããåæ°ã®å ¥åºåããã€ãšããŸãã æã ã®ç®æšã¯å¶åŸ¡å¯Ÿè±¡ã®åºå$y$ãšåç §å ¥å$r$ãšã®åå·®$e:=r-y$ãåãåããå¶åŸ¡å ¥å$u$ãçæããå¶åŸ¡åš$K(s)$ãå®è£ ããããšã§ãã ããã§ãå¶åŸ¡åš$K(s)$ã¯ä»¥äžã®èŠæ±ãæºããããšãæ±ããããŸãã å®å®åïŒ$G(s)$ãå®å®åããïŒæçãªå€ä¹±$d$ãåç §å ¥å$r$ã«å¯ŸããŠã$y$ãæçãšãªãïŒ è¯ãå¶åŸ¡æ§èœïŒ$e$ãã§ããã ãå°ãããã ããã¹ãå®å®åïŒ$G_A(s)$ãå®å®åããïŒåŸè¿°ïŒ 1ã ããéæããããã®æ¹æ³ã¯ããããããã®ã§ãæ¬èšäºã§ã¯è§ŠããŸããã 2ãéæããããã«ã¯ãå€ä¹±$d$ãåç §å ¥å$r$ã倧ããªå€ãæã€åšæ³¢æ°é åã§ãæ床é¢æ° $$ S(s) = (I+G(s)K(s))^{-1} $$ ãå°ããããããšãæ±ããããŸãã ããã§ãã"å°ãã"ãšã¯ãäŒéé¢æ°ã®æ倧ç¹ç°å€$\sigma_{max}(S(j\omega))$ãå°ãããšããæå³ã§ãã 3ã«ã€ããŠèª¬æããŸãã äžè¬ã«ãå¶åŸ¡å¯Ÿè±¡ãå®å šã«ã¢ããªã³ã°ããã®ã¯äžå¯èœã§ãããäœããã®äžç¢ºãããå«ããšèããã®ãèªç¶ã§ãã ããã¯äŸãã°ãçã®ã¢ãã«ã$G_A(s)$ãšãããš $$ G_A(s) = (I+L(s))G(s) $$ ãšè¡šãããšãã§ããŸãã ããã§ã$L(s)$ã¯ä¹æ³çäžç¢ºãããè¡šãäŒéé¢æ°è¡åã§ãããæ¢ç¥ã®$m(\omega)$ãšä»»æã®$\omega$ã«å¯Ÿã㊠$$ \sigma_{max}(L(j\omega)) < m(\omega) $$ ãªãé¢ä¿ãæãç«ã€ãšããŸãã ããã§ãç°¡åãªèšç®ãããçžè£æ床é¢æ° $$ T(s) = G(s)K(s)(I+G(s)K(s))^{-1} $$ ãä»»æã®$\omega$ã«å¯Ÿã㊠$$ \sigma_{max}(T(j\omega)) \le \frac{1}{m(\omega)} $$ ãæºããããšãã$G_A(s)$ã®å®å®æ§ã®å¿ èŠååæ¡ä»¶ãšããŠå°åºã§ããŸãã ...