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

There are well{known results demonstrating the convergence of such discrete{time games to their continuous{time counterparts (see, for instance, 11], 19]). Consequently, (38) implies jV";b;R;m (s; x)? V";b;R;m(s; x)j CT;D jx? xj 8 s 2 0; T]; x; x 2 BD (0); D< 1:^^^ Once one obtains the above Lipschitz bound in the space variable, one may obtain the bound in the time variable without resorting to the discrete{time setting. We will not include the details here since they are technical. However, we do note that one shows that there exist Ax; Cx such that for{optimal RT p p w, one has E( T? jwtj2 dt)]1=2 Ax+ Cx". Using this and the bounds on the rst and second derivatives of`;, one obtains the result. 2 an analogous bound in the discrete{time case. Speci cally, one may obtain bounds of the form E exp(ajyn j2) jFn?1] exp (ajyn?1j2(1+ K13 )+ K14] Fn?1 a.s. Using the nesting i h?X E exp jwnj2= E exp( jw1j2)E exp( jw2j2)E exp( jw3j2)E:::] jF2] jF1

Remark 3.5 As noted just above the previous lemma, a di culty arose in attempting to bound a term R of the form E exp( jwj2 dt)]. Interestingly, one may be able to use elements of the above proof to obtain

and (32), this may yield a bound of the desired form in the discrete{time case. However, one would need to prove some convergence of near-optimal discrete{time disturbances to near{optimal continuous{time disturbances in order to make use of this. Consequently, this approach was not attempted.

Lemma 3.6 Given any sequences fbng; fmng such that bn; mn ! 1, there exists a subsequence which we also subscript by n (to reduce notational complexity) such that e V";bn;R;mn ! V";R e uniformly on compact sets where V";R 2 C 1;2 (and the derivatives converge uniformly on compact sets as";R is a solution to e well). Further, V

0= Vs+ 2" 2

max ai;j (x)Vxi xj+ min w2<m f(x; u)+ (x)w] rV+`R (x; u)? 2 jwj2 u2U i;j=1 V (T; x)= R (x):

n X

2

(39)

By Lemmas 3.1, 3.2 and 3.4, V";b;R;m; V";bnn;R;mn; rV";bn;R;mn are all uniformly bounded s P on compact sets. Combining this with (27) implies that i;j=1 ai;j V";b;x;R;mn are also bounded uniformly xi n j on compact sets. Then following the method in 17], Appendix E (see also 33]), one obtains the result. 2 Proof.

Recall process (1) of Section 2. De ne the modi ed risk{sensitive cost and value#) ("Z T R (y"; ut) dt+ R (y" )";R(s; x; u )= E exp 1 J T" s` t and V";R (s; x)= uinfs" log J";R (s; x; u ): 2U 14

(40) (41)

上一页第14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