Project: Intelligent Coordination and Planning for Human-Robot Swarms

Key information:

Abstract: 

As robots become more prevalent and influential in our daily lives they will increasingly be required to cooperate, coordinate, and share experience without prior planning. This creates a demand for organisational structures with which diverse groups of robots can collectively plan, coordinate, and act. Emerging technologies such as Large Language Models, foundational approaches, and embodied robotic transformer policies provide the ideal basis for a domain-agnostic system which generalises across platforms and tasks. By employing language-based negotiation and planning between multi-robot teams and human operators, information can be shared and propagated throughout a dynamic swarm without requirement for a shared understanding of the environment; language provides this interface. Future robotic systems may feature end-to-end language, ranging from human instruction, to inter-robot planning, to the action selection of the robot. This project seeks to understand this oncoming trend and enable robots to share knowledge, understanding, and learning between each other without the need for common morphology or models. By extension, this research places high importance on responsible impacts and products by utilising natural language as a near-universal mode of control and interaction; humans, without need for expertise, can always interface with a language-based system