文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Finite Time--Horizon Risk Sensitive Control and the Robust Limit under a Quadratic Growth A
侵权投诉

Finite Time--Horizon Risk Sensitive Control and the Robust Limit under a Quadratic Growth A

The finite time--horizon risk sensitive limit problem for continuous, nonlinear systems is considered. Previous results are extended to cover more typical examples. In particular, the cost may grow quadratically, and the diffusion coefficient may depend on

The lower value and the upper value of the game are given by V (s; x):= 2inf(s) sup J(s; x; !]; !); !2B(s) W(s; x):= sup sup J(s; x; u; u]): 2 (s) u2U (s)

(43) (44)

In 31] it is proved that, under the current assumptions, for su ciently large

; V satis es the following Dynamical Programming Principle: (DPP) R

V (s; x)= 2inf(s) sup fI1+ V (s+; y(s+ ))g; !2B(s) 2

where I1:= ss+`(y( ); !]( ))? 2 j!( )j2d; for all 0 s s+ T; and V is continuous viscosity solution of (IE): This section and the following one will be concerned with the lower value; analogous results will hold e for the upper value as well. In the next theorem we will prove that the value function, V", of the risksensitive problem converges to that of the Robust control problem as" ! 0: This result is an immediate consequence of the stability of viscosity solutions with respect to locally uniform convergence of the DPE for the risk-sensitive problem to the equation (IE) as" ! 0: e Theorem 4.1 V" ! V uniformly on compact sets of s; T]<n: Proof.

Suppose without loss of generality that s= 0: From Lemmas 3.1, 3.5 and Corollary 3.8 it follows that for large enough and for any D> 0 there exist CT;D; MT;D> 0 (independent of") such that e e jV"(s1; x)? V" (s2; z)j C T;D jx? z j+ js1? s2 j] 8 s1; s2 2 0; T]; x; z 2 B D (0) (45) and e (46) jV" (t; x)j MT;D; 8 t 2 0; T]; 8x 2 B D (0): e So, by Ascoli-Arzela Theorem, for any compact K<n, there exists"n ! 0 such that V"n converges n by the following standard uniformly on 0; T] K: We construct a sequence converging on all 0; T]< e diagonal procedure. Let"(1) be such that V"n converges uniformly to U 1 on 0; T] B(0; 1); and de ne by n (m) subsequence of"(m?1) such that V"nm converges uniformly to U m on 0; T] B(0; m): e induction an"n n m= U m?1 on 0; T] B(0; m? 1); thus U(t; x):= U m (t; x); (t; x) 2 0; T] B(0; m) By construction U e is the uniform limit of V"nn on 0; T] K; for any compact K<n: From the stability of the viscosity solution with respect to the locally uniform convergence of the DPE for the risk-sensitive problem to the DPE for the Robust control problem as" ! 0; it follows that U is a viscosity solution of (IE). Moreover in 31] it is proved that under the current assumptions the lower value V is locally Lipschitz in the state variable uniformly in the time variable. Thus the uniqueness Theorem 5.1 implies that U= V: 2 (1) ( ) ( )

5 Uniqueness and Robust Control Value In this Section we will prove a uniqueness result for (IE) in the set of nonnegative continuous functions with locally bounded superdi erential in the x variable and growing at most quadratically. More precisely, given V: s; T]<n !<; we de ne x)+ Dx V (t; x)= p 2<n: lim sup V (t; y)? V j(t;? y? (y? x) p x j jx?yj!0 17 ( )

0;

上一页第17页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email