The project aims at developing and implementing a cognitive framework that would enable robots to operate robustly in human environments and to communicate effectively with people using natural language. To this end, the project can be divided into three main lines of work. First, we will develop methods for on-line incremental learning of objects with their properties and visual structure using natural language dialogue and visual perception. Second, we will adopt an appropriate knowledge representation structure and develop methods to add new knowledge and ground it against the robot's perceptual space. Third, we will develop algorithms for parsing natural language instructions into robot executable plans.
The three lines of work will be integrated sinergistically into a system capable of interactively and incrementally grounding tasks visually into a growing ontology.
- Humanoid Robots
- Machine Learning (ML)
- Natural Language Understanding (NLU)
- 3D Perception
- Visual Servoing
- Cognitive Systems