Autonomous Knowledge Acquisition and Semantic World Modelling for Language-Grounded Human-Robot Collaboration
Not stated
- Funding
- Self-Funded PhD Students Only
- Application deadline
- Year-round applications
About the project
About the Project Human-robot collaboration requires robots to perceive, understand and reason about the world in ways that support natural interaction with people. While recent advances in computer vision and vision-language models have significantly improved object detection and language grounding, most robotic systems remain limited to recognising predefined objects using static knowledge representations. They typically lack the ability to autonomously acquire new knowledge, update their understanding as environments change, and explain their decisions to human users. Enabling robots to continuously build and maintain semantic knowledge of their environment remains one of the key challenges in cognitive robotics and intelligent autonomous systems. This project aims to develop a novel framework for autonomous knowledge acquisition and semantic world modelling that enables mobile robots to learn, maintain and reason about their environment through multimodal perception and natural language interaction. Rather than relying solely on predefined object models and manually constructed knowledge bases, the proposed research will investigate how robots can acquire new concepts from human interaction and their own observations, constructing a semantic representation that evolves throughout their operation. Building upon preliminary work on language-grounded semantic object mapping, the research will investigate methods for integrating visual perception, natural language understanding, semantic mapping and symbolic knowledge representation within a unified architecture. The robot will associate object identities with spatial locations, semantic properties and contextual relationships, allowing it to understand spoken requests such as locating or retrieving an object. The expected outcome is a novel cognitive framework that enables robots to autonomously construct, maintain and reason over semantic knowledge acquired through perception and interaction. The research will contribute to the development of intelligent service and collaborative robots capable of continuously expanding their understanding of the world and interacting naturally with humans in different environments.