This article presents a dynamic knowledge graph approach to developing distributed self-driving laboratories (SDLs) within the World Avatar project. The goal is to create an all-encompassing digital twin of scientific research laboratories, enabling collaboration and resource integration across organizations. The authors employ ontologies to capture data and material flows in design-make-test-analyse cycles, using autonomous agents as executable knowledge components to automate experimentation workflows. Data provenance is recorded to ensure findability, accessibility, interoperability, and reusability. The framework is demonstrated through a real-time collaborative closed-loop optimization of an aldol condensation reaction between two labs in Cambridge and Singapore. The knowledge graph evolves autonomously to achieve the scientist's research goals, with the two robots effectively generating a Pareto front for cost-yield optimization in three days. The approach addresses challenges in resource orchestration, data sharing, and data provenance, offering a holistic solution for distributed SDLs. The study highlights the potential of dynamic knowledge graphs in enabling global collaborative research networks and advancing scientific discovery.This article presents a dynamic knowledge graph approach to developing distributed self-driving laboratories (SDLs) within the World Avatar project. The goal is to create an all-encompassing digital twin of scientific research laboratories, enabling collaboration and resource integration across organizations. The authors employ ontologies to capture data and material flows in design-make-test-analyse cycles, using autonomous agents as executable knowledge components to automate experimentation workflows. Data provenance is recorded to ensure findability, accessibility, interoperability, and reusability. The framework is demonstrated through a real-time collaborative closed-loop optimization of an aldol condensation reaction between two labs in Cambridge and Singapore. The knowledge graph evolves autonomously to achieve the scientist's research goals, with the two robots effectively generating a Pareto front for cost-yield optimization in three days. The approach addresses challenges in resource orchestration, data sharing, and data provenance, offering a holistic solution for distributed SDLs. The study highlights the potential of dynamic knowledge graphs in enabling global collaborative research networks and advancing scientific discovery.