Proceedings of ISP RAS


State prehistory for complex object estimation in a control system based on cases.

L.E. Karpov, V.N. Yudin.

Abstract

Building a control system for non-stationary objects is a subsequent step in developing an approach to controlling not fully described objects using Case-Based Reasoning. Initially this approach was offered in some earlier projects of the authors. The Case-Based Reasoning style of controlling complex objects shows us the cycle with three constituent elementary steps: estimating the object state, generating a controlling action, estimating the object state after that action. For non-stationary objects this traditional approach should be dramatically improved because now the state estimation cannot be done unambiguously. While we are performing these steps the object state may be changed spontaneously. In this case the action that was prepared beforehand becomes inapplicable. In this project the authors are trying to develop a method that should decrease the level of ambiguity in estimating the state object before and after the action. The previous object behavior is the source of additional information - that is the main idea of the method offered here. The authors have developed the technique of limiting the differential set. This technique allows us to fully represent all the previous states of the object starting from the very first step of control process.

Keywords

control system, Case-Based Reasoning, Data Mining, case base, non-stationary process, object behavior, measure of closeness, incompleteness of description, differential set of states, state prehistory.

Edition

Proceedings of the Institute for System Programming, vol. 24, 2013, pp. 437-446.

ISSN 2220-6426 (Online), ISSN 2079-8156 (Print).

DOI: 10.15514/ISPRAS-2013-24-19

Full text of the paper in pdf (in Russian) Back to the contents of the volume