Enhancing Semantic Coordination for Edge-AI Enabled Robotic Systems via Evolutionary Generative Adversarial Mechanism

Edge-AI Enabled Robotic Systems are increasingly deployed in dynamic, decentralized environments where real-time semantic coordination is essential for autonomous decision-making and collaboration. These systems depend on Ontology Matching (OM) to achieve consistent interpretation of heterogeneous and evolving knowledge sources. However, conventional OM methods typically rely on static Similarity Measures (SMs), expert-provided reference alignments, and fixed evaluation criteria, which limit their applicability in real-world robotic settings characterized by diverse ontologies and constrained computational resources. In this article, we propose a novel, adaptive, and unsupervised matching approach based on an Evolutionary Generative Adversarial Mechanism (EGAM). EGAM integrates a Genetic Programming (GP)-based similarity generator and a Genetic Algorithm (GA)-based discriminator within an adversarial co-evolutionary framework to construct and iteratively refine similarity measures without requiring ground-truth alignments. To mitigate evaluation bias, we introduce three lightweight, reference-free metrics that guide the evolutionary search process. This design enables EGAM to operate effectively under edge conditions while accommodating semantic variability across ontologies. We validate EGAM using the OAEI Conference dataset and a suite of real-world robotic ontology matching tasks. Experimental results demonstrate that EGAM outperforms state-of-the-art methods in both alignment accuracy and scalability, offering a resource-efficient, interpretable, and real-time solution for semantic coordination in next-generation edge-AI Enabled Robotic Systems.