Estimation of operational intention by utilizing self-organizing map, ontology, and Petri-net

This paper proposes a new approach for an estimation of user's intention using Self-Organizing Map (SOM), ontology, and Petri-net. In order to construct the intention estimator, the ontology and Petri-net are used to classify the operational intention. The SOM technique is used to form the relation between the intention modes and operational state. A kind of knowledge data base of user's intention was constructed by using the ontology, and the knowledge was utilized to build a task scenario. By using the Petri-net model, an event-driven scenario of the task is constructed and is utilized to improve the SOM. An experiment using a virtual transportation task simulator to verify the presented approach was performed, and the user's intention during the machine manipulation was estimated from the time-series data which was obtained through the experiment. Comparing the estimated intention with the other one extracted by human analysis, the effectiveness of the estimator was analyzed.