Parallel stochastic dynamic programming: Finite element methods (Q1194519)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Parallel stochastic dynamic programming: Finite element methods |
scientific article |
Statements
Parallel stochastic dynamic programming: Finite element methods (English)
0 references
27 September 1992
0 references
A multidimensional Markov dynamical system with continuous time is considered. The system includes a jump Poisson process and Gaussian motion as well as nonlinearities. A performance criterion of the system is introduced. It includes instantaneous cost and terminal cost functions. The general objective is to optimize the expected value of the performance criterion over a finite time horizon. This problem reduces to the problem of solving the Bellman functional equation of dynamic programming. To solve it the authors use the Galerkin approximation in the state space. This approach leads to a system of nonlinear algebraic equations. To this system the Crank-Nicholson predictor-corrector scheme is applied. Stability and convergence of the described method are verified. It is shown that the computer storage requirements for the proposed method are much weaker than those for the finite difference method.
0 references
parallel stochastic dynamic programming
0 references
finite element methods
0 references
stochastic control
0 references
stability
0 references
multidimensional Markov dynamical system
0 references
continuous time
0 references
jump Poisson process
0 references
Gaussian motion
0 references
Bellman functional equation
0 references
dynamic programming
0 references
Crank-Nicholson predictor- corrector scheme
0 references
convergence
0 references
0 references