Título: Neural Mechanisms for Real-Time Pattern Categorization in Robotics Applications
Autores: Gonçalves, Luiz M. G.; Distante, Cosimo; Oliveira, Antonio A. F.
Resumo: We present behavioraly active, self-growing neural mechanisms for pattern categorization based on multifeature extraction that can be used in (real-time) robotic applications. By using these mechanisms, robot agents are able to learn a pattern representation for different regions detected in a restricted environment and also to categorize already known (learned) patterns. As a practical result, the robotic agents, doted with an attentional mechanism, are able to perform a visual monitoring task of their underlying space. That is, they are able to construct attentional maps of the environment, and to keep the maps consistent with a current perception of the scene dealing in an efficient manner with new or already known pattern representations.
Código DOI: 10.21528/CBRN2001-097
Artigo em pdf: 5cbrn_097.pdf
Arquivo BibTex: 5cbrn_097.bib