Autonomous Feature-Based Exploration using Multi-Sensors

In this paper, we present an algorithm for feature based exploration of a prior unknown indoor environment. The mobile robot Odete equipped with two different types of sensors is used. Employing different sensor capabilities, 3D features are detected autonomously very quickly. The main contribution of the paper is to let the robot move to a new unexplored place, based on some prospect values of this place. One of these prospect values is the natural feature quality available at that place, whereas the second prospect value is the navigation cost to it. The vision-based feature extractor is demonstrated, showing how it is self motivated to search for an interesting region using a segmentation and detection process. The obtained region is classified into known objects, and the relative location information for the selected object is memorized. The mobile robot plans its motion such that it continuously builds up a geometrical model of the features, taking into consideration the cost of reaching a target point. The target point is chosen depending on the number of the features and their visibility that a robot can cover with its sensors upon reaching that location. The captured information from these sensors is transformed into different representations in order to achieve the task