Information and Computation
Quantization effects on the equilibrium behavior of combined fuzzy cognitive maps: Research Articles
International Journal of Intelligent Systems
Application of the Holonic Approach in Distributed Control Systems Designing
HoloMAS '07 Proceedings of the 3rd international conference on Industrial Applications of Holonic and Multi-Agent Systems: Holonic and Multi-Agent Systems for Manufacturing
Fast universal coding with context models
IEEE Transactions on Information Theory
A universal finite memory source
IEEE Transactions on Information Theory
The context-tree weighting method: basic properties
IEEE Transactions on Information Theory
| Hi-index | 0.00 |
This paper presents a reconfigurable multi-agent system (MAS) applied to distributed control system (DCS). Agents co-operate, evaluating the system on different levels of information abstraction. Context-based statistical remote expert agent or local supervisory agent are used to evaluate control agent performance. Using expert-selected period of good performance the reference distribution function is proposed. Periodically, for any monitoring period, a sample of observations is taken for local or remote system performance monitoring. Because evaluation may also be carried out remotely two cases should be considered. Remote expert observes changes of parameters that come from process performance degradation. Second case refers to communication problems when data transmission is corrupted and can not be used for system evaluation. Because of that application, context model is necessary to inform the remote expert about transmission quality. For evaluation of transmission channel, the idea of a context tree is utilised. Number of nodes and leaves taken into considerations depends on the expert's knowledge.