STATE INCREMENT DYNAMIC PROGRAMMING APPLIED TO STOCHASTIC AND ADAPTIVE OPTIMIZATION PROBLEMS.
Abstract
Dr. R. E. Larson has proposed an optimization technique, state increment dynamic programming, which reduces the fast digital computer memory requirement of conventional dynamic programming while retaining its many desirable features. The primary modification in state increment dynamic programming is that, instead of a fixed time of application of control as in conventional dynamic programming, the duration of application of control is selected as the time such that no state changes by more than one increment. This guarantees that the state resulting from the application of a control lies in a small neighborhood of the original state. Consequently, the entire region of state space is divided into smaller units called 'blocks.' The smaller amount of information contained in one of these blocks, as compared with the entire region of state space, accounts for the reduction in the fast memory requirement. State increment dynamic programming is applied to stochastic optimization problems in which the stochastic variable is a set of values each with a known discrete probability. The equations are presented in a form suitable for a digital computer solution.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1966
- Accession Number
- AD0639702
Entities
People
- John C. Koger
- Robert E. Heath Ii
Organizations
- Air Force Institute of Technology