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

8x 2<n; (x) 0; 8x 2<n; 2); (A4) iii) j (x)j C(1+ jxj 8x 2<n;>> iv) j (x)? (y)j C(1+

jxj+ jyj)jx? yj; 8x; y 2<n;>>: v) j (x)? (y)j MR jx? yj; 8jxj; jyj 2 fz: (z) Rg; 8R< 1; where L; L 2<; C; K; M; 2 0; 1), MR< 1 for all R< 1, and L(<m;<n) appearing in (A2:i) C;

8> i)>>> ii)<

2 C 2(<n;<);

j

xx(x)j

represents the space of n m matrices. Note that (A3:iv), (A3:v), (A4:iv) and (A4:v) are used only to obtain the results in Section 3. This is also true of the requirement of continuous, bounded second{ derivatives in (A3:i) and (A4:i). Also note that (A2:iv), (A2:v) are not needed for the results on the Robust problem of Section 5. It seems that many of the above smoothness assumptions which are required in the pre{limit analysis but not in the limit uniqueness result might be removable, however the pre{limit analysis is already quite technical, and so no attempt is made to reduce the assumptions further. There will be no further assumptions. e In Section 3, it will be shown that there exists a classical solution, V" 2 C 1;2 to partial di erential equation (PDE) 0= Vs+ 2" 2 n X i;j=1

V (T; x)= (x)

ai;j (x)Vxi xj+ 21 2 rV T a(x)rV+ min f(x; u) rV+`(x; u)]; 0 s T u2U

(4)

: where a= T . For the purposes of this section, assume that the classical solution exists, and we postpone the proof of this fact to the next section. Note that 1 rV T arV= max ( w) rV? 2 kwk2: (5) w2<m 22 2 De ne a measurable (see for instance 17]) feedback control e u(t; x) 2 argmin f(x; u) rV" (t; x)+`(x; u)]:

(6)

Using 41], one proves the existence of a strong solution, y", to (1) with feedback u (see 33]). (Also see 3] for a quicker proof of a weak solution.) Then de ne the Ft{progressively measurable control u given e: by ut(!)= u(t; y" (!)) for all ! 2 . e t The rst lemma is rather easily proved.

Lemma 2.1 Proof.

" log J" (s; x; e) V"(s; x) u e By Ito's rule and (1), Z

8 s 2 0; T]; 8 x 2<n: n X Z

" V" (yT )? V" (x)=

Th

s

f(y"; ut) t e

r

e e V"+V"+

t

2 2 i;j=1 Z

"

ai;j (y" )Vx"i xj t e

i

dt+

Tr

s

" (rV" )T e 2

dBt:

Using (4) and (6) one nds Z e`(y"; ut) dt+ (y" )= V" (x)+ t e T s

T

r

"

2 s

T

Z T e e e (rV")T dBt? 21 2 (rV" )T arV" dt

s

上一页第4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