摘要
The traditional linear quadratic control can be solved elegantly via the classical algebraic Riccati equation. Unfortunately, this Riccati approach is rather limitation to handle the indefinite linear quadratic control, since the inverse(generalized inverse) of indefinite matrix is hard to express. The semidefinite programming(or linear matrix inequality) based approach could tackle indefinite linear quadratic control in a few cases, however, the complexity of solving the semidefinite programming is impractical for real-time control, especial in large scale problem. In this paper, we discuss two classes of indefinite linear quadratic control, the corresponding conic quadratic formulations are derived. An attractive feature of the proposed formulations is the numerical tractability, especially when compared to approaches based on semidefinite programming.
The traditional linear quadratic control can be solved elegantly via the classical algebraic Riccati equation. Unfortunately, this Riccati approach is rather limitation to handle the indefinite linear quadratic control, since the inverse(generalized inverse) of indefinite matrix is hard to express. The semidefinite programming(or linear matrix inequality) based approach could tackle indefinite linear quadratic control in a few cases, however, the complexity of solving the semidefinite programming is impractical for real-time control, especial in large scale problem. In this paper, we discuss two classes of indefinite linear quadratic control, the corresponding conic quadratic formulations are derived. An attractive feature of the proposed formulations is the numerical tractability, especially when compared to approaches based on semidefinite programming.
引文
[1]Kalman R.E.,Contribution to the theory of optimal control,Bol.Soc.Mat.Mexicana,2(5),102-119,1960.
[2]Ferrante,A.,Ntogramatzidis,L.,The generalized discrete algebraic Riccati equation in linear-quadratic optimal control.Automatica,49(2):471-478,2013b.
[3]Ferrante,A.,Ntogramatzidis,L.,A reduction technique for discrete generalized algebraic and difference Riccati equations.Linear and Multilinear Algebra,62(11):1460-1474,2014.
[4]Berkovitz L.D.,Optimal Control Theory.New York:Springer,1974.
[5]Boyd S P,El-Ghaoui L,Feron E,Linear Matrix Inequalities in System and Control Theory,SAM[J].Studies in Applied Mathematics,15(6),1994.
[6]Lewis F.L.,Syrmos V.,Optimal control,New York:John Wiley&Sons,1995.
[7]Ran,A.C.M.,Trentelman,H.L.,Linear quadratic problems with indefinite cost for discrete time systems.SIAM Journal on Matrix Analysis and Applications,14(3):776-797,1993.
[8]Hassibi,B.,Sayed,A.H.,Kailath,T.,Indefinite-quadratic estimation and control,a unified approach to H2and H∞theories,Philadelphia:SIAM.,1999
[9]Bilardi,G.,Ferrante,A.,The role of terminal cost/reward in finite horizon discrete-time LQ optimal control,Linear Algebra and its Applications,425:323-344,2007.
[10]Vandenberghe,L.,Boyd,S.:Semidefinite programming.SIAM,38:49-95,1996.
[11]Yao D D,Zhang S,Zhou X Y.,A primal-dual semi-definite programming approach to linear quadratic control[J].IEEE Transactions on Automatic Control,46(9):1442-1447,2001.
[12]Nesterov,Y.,Nemirovsky,A.,Interior-Point Polynomial Methods in Convex Programming,SIAM,Philadelphia,1999.
[13]Yao D D,Zhang S,Zhou X Y.,Stochastic Linear-Quadratic Control via Semidefinite Programming[J].SIAM Journal on Control&Optimization,40(3):801-823,2001.
[14]Uhlig F.,Definite and semidefinite matrices in a real symmetic matrix pencil,Pac.J.Math.,49(2),561-568,1972.
[15]Bertsekas D.,Dynamic Programming and Optimal Control.Nashua,NH:Athena Scientific,2005.
[16]Aharon Ben-Tal,Dick den Hertog,Hidden conic quadratic representation of some nonconvex quadratic optimization problems,Math.Program.,Ser.A,143:1-29,2014.