Distributed and incremental visual object categorization for humanoid platform NAO

Abstract

Robotics is an essential component of current and future technological existence of mankind. Interaction between human and technology such as robots and their coexistence with humans will be extremely important. In fact, a fully embodied intelligence is one version of the future but another option is distributed intelligence when a part of intelligence is "somewhere in the Cloud", and the second part is on the robot itself. Certainly there are several questions about connectivity and reduced activity in case of off-line robot life comparing to life in on-line robot existence. Meanwhile, a notion of Cloud Robotics came out, and it brings new challenges and endless possibilities for the future. The paper gives theoretical and experimental research about utilization of selected Neural technology specifically modified MF-ARTMAP neural network for object categorization. We used the categorization test-bed a Humanoid platform NAO connected to Global framework (MASS), which categorize in distributive and incremental approach objects. This approach creates a power of sharing knowledge among other robots and building collective intelligence. Papers gives the results of a pilot study and refers about preliminary results on different experimental data with building a database of objects on off-robot ecosystem. The system also works with fuzzy logic relations among objects and maintain a list of statements about the relationship between objects seen by the robot. © 2013 Springer-Verlag.