Credit: CTL
Dr. Martin Serrano, Amelie Gyrard, and Eoin Jordan, in collaboration with NIST researchers Eugene Song, Tom Roth, and David Wollman, received a best paper presentation award at the 2024 IEEE IECON international conference for the team’s research paper, titled “Semantics for Enhancing Communications- and Edge-Intelligence-enabled Smart Sensors: A Practical Use Case in Federated Automotive Diagnostics.” In his presentation, Dr. Serrano explained the advantages of semantics and ontology engineering and its relevance for enabling semantic communications in edge AI-enabled smart sensors, and its use in federated automotive systems.
Dr. Serrano introduced the visionary data continuum for IoT data including interpreting, understanding, and visualizing IoT data from various sensor devices to applications, and he addressed the new challenges of IoT for semantic communication to collect, exchange, and share IoT data. Dr. Serrano explained how semantic Web technologies based on linked data and data interoperability principles can be used to enable efficient stream processing, data management, and intelligent applications.
Dr. Serrano then described the paper contributions including extraction of domain knowledge (e.g., dictionary, rules to interpret data, and the use of ontology information) and use of this knowledge in automotive applications. Dr. Serrano explained how linked Open Vocabularies for Internet of Things (LOV4IoT) and the Machine to Machine for Measurement data (M3) framework can be used to combine and infer domain knowledge for applications. He described a practical use case of AI-enabled smart sensors for automotive diagnoses using an audio event dataset classifier (YAMNet) and Tiny Machine Learning (TinyML), and he demonstrated a prototype implementation on edge devices including its practical use for event detection in Automotive Preventive Maintenance.
Dr. Serrano concluded by describing planned future work including the design of an AI-enabled smart sensor architecture, other applications and uses of the introduced work in different domains, AudioSet extensions, and semantic model validation including the need for cross-domain knowledge testing using the presented edge AI approach.
