Semantic Stability: A Missing Link between Cognition and Behavior

Main Article Content

Mahmudul Hasan
Md. Rokunuzzaman, Barun Kumar Das
Asma-Ul-Husna, Md. Abdul Kader

Abstract

This paper explains the concept of semantic stability as a missing link between cognition and behavior for multi-robot cooperation with distributed sensing. In cooperation perspective when the cognition of robots is different, there need to interpret the situation by selecting a common Region of Interest. Interpretation of situation by observing visual information is still challenging for the artificial agents. The reason behind this difficulty is to find the cognitive boundary by which the situation can be described efficiently. Since this boundary is a part of cognition, therefore this research relates this boundary with region of interest (ROI) to visualize this concept. In this regard, this boundary contains semantic relevant objects which build a connection between visual objects and their semantic interpretation by semantic labelling. This paper addresses various related technical issues to emphasize the emergence of this concept. A formal expression for the semantic stability concept has been proposed in the current work and applies it to determine the cognitive boundary by selecting the best ROI from the visual scenes. However, selecting of best ROI is meaningless, since the cognition of individual robot is different. Therefore, for co-operation, best ROI needs to be shared between robots. In this research, ROI sharing gives various behaviors which are guided by semantic stability concept. In this respect, semantic stability acts as a missing link between cognition and behavior by selecting meaningful ROI for Cognition and decision making to perform actions which stands as Behavior in a cooperation task. Several complex scenes are tested to illustrate the applicability of the proposed algorithm at the end of the paper.


Keywords:Semantic Stability, Region of Interest (ROI) sharing, Cognition, Multi-robot Co-operation

Downloads

Download data is not yet available.

Article Details

Section
Articles