随机控制的极大值原理及其应用.docVIP

  Maximum Principle of Stochastic control problems And it’s Application 5 10 15 20 25 30 35  Abstract: A maximum principle of stochastic control is present for solving dynamic optimization problems involving stochastic differential equation .It is an alternative to dynamic programming. As a generalization of Pontryagins maximum principle to stochastic control models in Economics and management it avoids having to solve the bellman equation for the value function. Its analytical convenience is illustrated by application to classic problems of investment decision and stochastic project with optimum storage and productivity. Key words: stochastic control; product project; investment decision; dynamic programming 0 Introduction In economics, finance and management science, many problems are concerned with stochastic optimal control. In general. the optimal state estimation can be obtained for the linear systems with noise and noisy measurements. The white noise process can also be defined for continuous time systems for obtaining the Kalman-Bucy filter [1, 2]. For stochastic linear quadratic optimal control problems, the separation principle may be used to obtain the optimal estimate of the state first and then use it to determine the optimal feedback control formula [3]. For more general stochastic optimal control problems, for instance, stochastic production planning and stochastic optimal decision of investment etc. [4, 5], and the separation principle does not hold. To find solutions for these problems requires knowledge of stochastic differential equation [6, 7] and partial differential equations [8, 9]. The solution method makes use of the Hamilton-Jacobi-Bellman theory that is extremely complexity and it is difficult to solve to obtain analysis solutions and can only gained numerical solutions [10, 11]. A maximum principle that can be used to solve general stochastic control problem has been presented in this paper. It develops the applied area of Pontryagin