Adaptive Navigation for Mobile Robots with Object Recognition and Ontologies

The ability of a mobile robot to navigate in unknown and dynamic environments depends on some issues, like localization, sensors data processing (or mapping) and decision making. According to the environment characteristics, the robot needs to make different decisions, aiming to finish its mission, to preserve its physical integrity and the environment integrity. This paper presents an intelligent navigation system based on objects recognition and its semantics. The proposed system permits a mobile robot to assume different behaviors according to the recognized objects and their properties. Furthermore, the mobile robot can predict some behavior inherent to the recognized objects, helping it to interact coherently with its environment and objects. In order to achieve this purpose, the presented system uses the SURF algorithm to perform object recognition over Kinect images and ontologies techniques to define an object and its properties. At the end of this paper some preliminary experiments are shown with real data.