Iterative learning control for discrete-time systems with exponential rate of convergence
Citations
2,645 citations
Cites background from "Iterative learning control for disc..."
...The above optimization problem can be solved in a different way leading to a combination of optimal state feedback control and current-iteration ILC [100], [102]....
[...]
...Minimizing the cost criterion with respect to uj+1 [100], [101] yields the optimal Q-filter FIGURE 4 ILC iteration dynamics....
[...]
[...]
1,829 citations
1,417 citations
451 citations
Cites background from "Iterative learning control for disc..."
...Amann et al. (1996) transformed (4) to a causal form by borrowing the idea from the solution of the "nite-time quadratic optimal tracking problem....
[...]
...In the unconstrained, deterministic setting, Amann et al. (1996) derived a noncausal input updating law u k "u k~1 #R~1GTQe k (4) from LJ k /Lu k "0, while Lee et al. (1996) obtained u k "u k~1 #(GTQG#R)~1GTQe k~1 (5) which is indeed a rephrasing of (4) in a pure learning form....
[...]
397 citations
References
28,888 citations
"Iterative learning control for disc..." refers methods in this paper
...Using e = r Gu gives the tracking error update relationek+1 = (I +GG ) 1ek 8k 0 (17) and the recursive relation for the input evolutionuk+1 = (I +G G) 1(uk +G r) 8k 0 (18) This last relationship is a form of Levenberg-Marquardt [14] or modified Newton iteration which is familiar in the context of numerical analysis [18], particularly the least-squaresfitting of parameters appearing nonlinearly in models, but in this case it is used for a dynamical system....
[...]
...A formal examination of general robustness issues was not included, but the experience in numerical analysis with the related Levenberg-Marquardt method and results of Iterative Learning Control simulations indicate that the algorithmpossesses robustness to a useful degree....
[...]
...Using e = r Gu gives the tracking error update relation ek+1 = (I +GG ) 1ek 8k 0 (17) and the recursive relation for the input evolution uk+1 = (I +G G) 1(uk +G r) 8k 0 (18) This last relationship is a form of Levenberg-Marquardt [14] or modified Newton iteration which is familiar in the context of numerical analysis [18], particularly the least-squaresfitting of parameters appearing nonlinearly in models, but in this case it is used for a dynamical system....
[...]
19,881 citations
5,667 citations
Additional excerpts
...The controller on the (k + 1)th trial is obtained with vector differential calculus from the required stationarity condition 12 @Jk+1 @uk+1 = GTQek+1 + R(uk+1 uk) = 0 (15) Since R(t) > 0 8t guarantees the existence of the inverse, the optimal control input is uk+1 = uk + R 1GTQek+1 8k 0 (16) The learning controllerR 1GTQ is equivalent to the adjoint operatorG ofGwith respect to the weighted inner products (12) and (13) [13]....
[...]
[...]
3,133 citations