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

(DPE) in the in nite time{horizon case which is not a problem in the nite time{horizon case. Th

at is, there may be an in nite number of viscosity solutions and multiple classical solutions (when they exist) to the DPE. (This is in addition to the multiplicity incurred by scaling with an additive constant.) This question is addressed in 29], 30]. One approach to the nite time horizon problem is to prove that the value function is a viscosity (or classical) solution to the associated DPE for the risk sensitive problem. Let the parameter associated with risk{sensitivity be" (described more accurately below). This parameter will also be used scale the di usion term. One wishes to show that as"# 0 (i.e. as the problem becomes in nitely risk averse), that the value function of the risk sensitive problem converges to that of a Robust (H1 ) control problem. It is easily seen that the DPE for the risk sensitive problem formally converges to the DPE for the Robust control problem as"# 0. One may also show that the value of the Robust control problem is a viscosity solution of this limit DPE. If one proves uniform boundedness and equicontinuity results for the pre{limit value functions, then by the stability of viscosity solutions, these converge (subsequentially) to viscosity solutions of the limit DPE. A key step is to prove a uniqueness result for viscosity solutions of this limit DPE. Then, one can assert that the solutions of the pre{limit DPE (the risk sensitive values) converge to the unique solution of the limit DPE which must also then be the value of the limit Robust control problem. One of the key di culties was that the viscosity solution uniqueness results did not cover the limit DPE's except under rather strict conditions such as globally Lipschitz conditions on the cost and dynamics. In 32], viscosity solution uniqueness results were extended to cover certain cases where the cost could grow quadratically. However, it was still required that one prove a certain Lipschitz condition with Lipschitz constant growing linearly with the norm of the state, uniformly for the pre{limit solutions as well as the limit. For unknown reasons, this has been di cult to prove in the case where the di usion coe cient depends on the state. Some recent results (obtained independently and concurrently) 2], 21] have extended the uniqueness results in ways that do not require this Lipschitz condition. These results allow one to extend the previous risk sensitive limit results to a larger class which subsumes a large variety of LEQG problems (but not all). This extension is the subject of this paper. Consequently, the uniqueness result is a central topic of the paper. Some other technical di culties arise in this case as well. In particular, there is some di culty with a Novikov condition that was not present previously. Some di culties also arise in the proof of the equicontinuity of the value functions. As intimated above, one would expect a Lipschitz condition with linear growth of the Lipschitz coe cient (such as in the case of a quadratic function) to

hold uniformly in". Although, this seems di cult to prove, it is su cient, with the uniqueness result that we will use, simply to prove equicontinuity. In Sections 2 and 3, a representation result will be obtained that states that any classical solution of the risk sensitive DPE must be the value function of the risk sensitive problem. The assumptions will also be listed in that section. Further, in Section 3, the existence of the above classical solution will be proved, and the uniform boundedness and equicontinuity results will be obtained along the way. In Section 4, we brie y state the limit results. Of course these results require the uniqueness result discussed above. This uniqueness result is stated and proved in Section 5. At the same time, it is shown that this unique solution is the value of the Robust control problem, thus making the connection between the risk averse limit of the risk sensitive problem and the Robust control problem.

上一页第2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