A Scalable Method for Robot Object Recognition Using Augmentable Ontology

An intelligent robot must have knowledge acquired and maintained either locally or remotely. When a robot needs to recognize an object, it would make a decision based on the information in the knowledge base. Conventional ways to construct the knowledge models, especially during the object recognition process, are quite rigid and not scalable. In this paper, we propose a new flexible recognition model using the ontology concept. Ontology provides platform-independent representation for data and schema for the relationship among data. By using this, the numerical data collected through the sensory units of the robot is transformed into more semantic and meaningful form, and this semantic information helps to determine what object has been detected. This ontologybased approach contributes to enhance the object recognition performance, especially in an environment where unexpected situations are encountered. We implemented a knowledge management system for intelligent robots, and the results of testing ontology models in this environment are satisfactory.