Towards the vision of a social robot in every home: A navigation strategy via enhanced subsumption architecture

In this thesis, we report the studies undertaken in the design and implementation of a behavioristic navigation system for social robots with limited sensors to be deployed in family homes. The project was completed in four phases. Each phase of the project was independently evaluated in virtual or real-time implementation on the NAO humanoid robot. In the first phase of this research study, we address the problem of indoor room classification via several convolutional neural network (CNN) architectures. The main objective was to recognize different rooms in a family home. We also propose and examine a combination model of CNN and a multi-binary classifier referred to as Error Correcting Output Code (ECOC). In the second phase, we propose a new dataset referred to as SRIN, which stands for Social Robot Indoor Navigation. This dataset consists of 2D colored images for room classification (termed SRIN-Room) and doorway detection (termed SRIN-Doorway). The main feature of the SRIN dataset is that its images have been purposefully captured for short robots (around 0.5-meter tall). The methodology of collecting SRIN was designed in a way that facilitated generating more samples in the future regardless of where the samples have come from. In phase three, we propose a novel algorithm to detect a door and its orientation in indoor settings from the view of a social robot equipped with only a monocular camera. The proposed system is designed through the integration of several modules, each of which serves a special purpose. Finally, we report an end-to-end navigation system for social robots in family homes. The system combines a reactive-based system and a knowledge-based system with learning capabilities in a meaningful manner for social robot applications.