engleski

Context driven model for simulating human perception - a design perspective

This paper presents ongoing research about a cognitive model used to simulate processes of human perception. The model could be understood as a context-to-data interpreter that transforms high- level context into low-level data, allowing machines to make context- based decisions. The design philosophy is focused here on the effects of perception without defining the phenomenon itself. The results are used as the basis for a novel approach to robot reasoning. Compared to conventional approaches where robots are preprogramed to react on a finite number of occurrences, the proposed approach can enable adaptation to changes in a way similar to human adaptive mechanisms. The first part of the model is used to collect significant environmental information. The second part generates possible robot reactions respecting the semantic domain description. The model contains also a probabilistic component ensuring a single solution in a given context.

Cognition; perception; ontology; probabilistic robotics; model

nije evidentirano