CAREER: Internet of Foundation Models (IoFM): Expanding the Frontiers of Distributed Learning in Heterogeneous Networks

This project aims to advance AI machine-learning tools for future wireless networks by introducing a transformative paradigm called the Internet of Foundation Models (IoFM). Modern wireless networks are evolving into intelligent infrastructures that generate vast amounts of distributed data from Internet of Things (IoT) devices. This project seeks to harness this data to enable real-time decision-making, situational awareness, and autonomy in next-generation wireless systems. By leveraging multi-modal multi-task foundation models (M3T FMs), which can process diverse data types and perform multiple tasks simultaneously, IoFM envisions a globally distributed ecosystem that integrates these models across cloud, edge, and device networks. The project promises to enhance the efficiency, scalability, and privacy of AI applications in wireless systems, with potential societal benefits such as improved healthcare, smarter cities, and more sustainable technology systems. Additionally, the project includes a robust education and outreach plan to train the next generation of engineers and researchers, engage K-12 educators, and foster interdisciplinary collaboration, ensuring long-term societal and economic impact. The research of this project focuses on developing scalable, resource-efficient, and privacy-preserving methods for deploying multi-modal multi-task federated foundation models (M3T FedFMs) across heterogeneous wireless networks. The research is structured around four dimensions of heterogeneity: data, orchestration, model, and environmental variability. The project is divided into three research thrusts: (1) establishing theoretical foundations for modeling data and model heterogeneity, (2) designing scalable AI learning architectures for hierarchical and decentralized fog networks, and (3) optimizing cross-layer communication, computation, and storage resources for efficient deployment. The methods include modular AI learning algorithms, hierarchical and device-to-device (D2D) AI learning protocols, and advanced AI optimization techniques such as reinforcement learning and Lyapunov optimization. The project will also feature extensive experimental validations using simulations, testbeds, and real-world datasets. By addressing the challenges of deploying M3T FedFMs in resource-constrained environments, the project aims to advance the state of distributed machine learning, wireless network optimization, and AI applications. The outcomes will contribute to the development of intelligent wireless systems capable of supporting diverse applications, from autonomous vehicles to healthcare, while ensuring privacy and resource efficiency. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.