Plan Reconsideration in Hybrid Agent System

Plan Reconsideration in Hybrid Agent System

Paper in proceedings outside WoS and Scopus

bThere is discussed an idea of agent's plan reconsideration in this paper. The approach is done by a hybrid agent architecture where plan completed by a deliberative unit is checked by another reactive part. The reactive part could be of pure rective kind that means that action is selected only on basis of actual situation of the environment, or it also can be some learning predictive system (in this case Back Propagation Neural Net). Plan reconsideration depends of correlation between the established plan and actions chosen by the reflexive part on each step of the execution. Architecture of such agent is proposed in the text.

bThere is discussed an idea of agent's plan reconsideration in this paper. The approach is done by a hybrid agent architecture where plan completed by a deliberative unit is checked by another reactive part. The reactive part could be of pure rective kind that means that action is selected only on basis of actual situation of the environment, or it also can be some learning predictive system (in this case Back Propagation Neural Net). Plan reconsideration depends of correlation between the established plan and actions chosen by the reflexive part on each step of the execution. Architecture of such agent is proposed in the text.

Agent System, Planning, Neural Network

Agent System, Planning, Neural Network

ZBOŘIL, F.; ZBOŘIL, F.

2011

08.09.2003

Marq software s.r.o.

Ostrava

80-85988-88-7

Proceedings of XXVth International Autumn Colloquium ASIS 2003

315

320

6