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

3 Existence and Uniformity The representation results of the previous section required the existence of a C 1;2 solution to (4). That will be obtained in this section. Also, in order to prove the risk{averse limit result to follow, we will e need continuity and local boundedness estimates on V" inde

pendent of"> 0. These will be obtained at the same time as the existence result. Lastly, we will obtain the promised improvement over Lemma 2.3. These results will all be obtained through the stochastic game representation. Before proceeding with the technical details to follow, we give some formal indications of the general concepts. Using convex duality, one may rewrite DPE (4) as max ai;j (x)Vxi xj+ min w2<m f(x; u)+ (x)w] rV+`(x; u)? 2 jwj2 u2U i;j=1 V (T; x)= (x): 0= Vs+ 2" 2 n X

2

(14)

In this form, it is intuitive that this PDE might be associated with the stochastic di erential game with dynamics " dyt= f(yt"; ut)+" ys= x

(yt" )wt] dt+

r

" (y" ) dB t 2 t

(15) #

and payo Jg" (s; x; u; w)= E

"Z

T

s

"`(yt; ut)?

2" 2 jwtj dt+ (yT ):

2

In this game, w is the control for the opposing (disturbance) player who is trying to maximize this payo which we are trying to minimize. The de nitions of value for such a game 19] are discussed further below. The connections between such games and their corresponding DPE's (now second{order Isaacs equations) are not known for the class of systems considered here, so it will be necessary to consider some more restrictive assumptions, and then to relax these assumptions to obtain the above game and DPE. One technique that is sometimes useful for games associated with H1 control is to obtain an L2{ bound on"{optimal controls, w, for the disturbance player 31], 29], 30]. This is then used to obtain continuity estimates on the value function. The simplest analogue is to prove a bound of the form kwkL ( s;T] ) M< 1 for"{optimal w. Such a bound is not di cult to obtain in this case, but it does not yield the equicontinuity estimates for the value needed for the proofs to follow. Instead, one can obtain an almost sure bound. This bound is rather technical due to the presence of the Brownian motion with its unbounded variation. We now proceed with the analysis. We rst work with a system where the dynamics and cost functions are all bounded. Let (16)`R (x; u)=`(x; u) if`(x; u) R R otherwise, R (x) if (x) R R (x)= (17) R otherwise f(x; u) if jf(x; u)j b f b (x; u)= b f (x;u) otherwise. (18) jf (x;u)j 2

上一页第7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