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

Lemma 3.4 For su ciently large (i.e. but independent of b; R; m;") such that

), there exists C T;D< 1 (depending on T; D 2 0; 1)

s^^^ jV";b;R;m (s; x)? V";b;R;m(^; x)j C T;D jx? xj+ js? sj] Actually CT;D is linear in D for the space estimate.

8 s; s 2 0; T]; x; x 2 BD (0):^^

Proof. We rst obtain the Lipschitz estimate in the space variable. This requires the discrete game approximation. Once one has this, the estimate in the time variable follows without having to use discrete approximations. De ne the discrete approximation as follows. Let the number of time{steps be N, and the step{size: is then= T=N. We assume 1 freely throughout the proof. Let Bn be a sequence of IID m{ dimensional normal random variables with mean 0 and covariance Im . Let the Bn be adapted to the sequence of{algebras Fn where F0 is trivial. Let the disturbance at step n be wn which is measurable with respect to Fn?1. We suppress the appearance of the control in this discrete approximation estimates, since the proof is already quite long, and it should be clear that the controller does not introduce any additional problems. The dynamics are given by

yn= yn?1+ f(yn?1)+ (yn?1 )wn]+" 2 (yn?1 )Bn:

s

(29)

Let W N be the set of sequences of L2 disturbances, fwngN=1, of length N where each wn is Fn?1{ n measurable. For any w 2 W N and initial state, y0= x, de ne the payo J N (x; w)= E and value "

N X i=1

`(yn?1 )? w2W

2 2 jwnj]+ (yN )

2

#

V N (x)= supN J N (x; w):

We will use dynamic programming to work back from our assumed Lipschitz continuity estimate on to obtain the estimate on V N . Let V N (xN )= (xN ) N h? i 2 J N?1(xn?1; vn )= E`(xn?1)? 2 jvnj2+ V N (yn ) n n N N V n?1(xn?1)= supm J n?1(xn?1; vn ) 8 n= 1; 2;:::N vn 2<

where yn is given by (29) with yn?1= xn?1 and wn= vn where xn?1; vn are deterministic. Then, of course, V N (x)= V N (x): (30) 0 First we obtain a bound on jvN j depending on jxn?1j. Given vN, let yN= xN?1+ f(xN?1 )+ (xN?1 )vN]+" 2 (xN?1 )BN 0 yN= xN?1+ f(xN?1 )]+ s s

(31)

"

2 (xN?1 )BN:

上一页第11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