A Survey of FPGA Integration Techniques in IoT Infrastructure for AI/ML Tasks

In this work the integration of Field Programmable Gate Arrays (FPGAs) in Internet of Things (IoT) infrastructure for Artificial Intelligence (AI) and Machine Learning (ML) tasks was analyzed. Existing approaches and methods for FPGA integration were investigated including challenges posed by limited computational resources, power constraints, and real-time processing requirements at the edge. A review of the state-of-the-art in FPGA technology applied to IoT was done, highlighting the advantages of FPGA reconfigurability and high performance for AI/ML workloads. The limitations and obstacles in FPGA integration across different layers of IoT infrastructure were also explored. The proposed techniques aimed to optimize FPGA deployment for intensive computational tasks, enabling real-time decision-making, reducing latency, and enhancing data privacy by minimizing the need for centralized cloud data transfer. Additionally, the benefits and drawbacks of these techniques were evaluated, and recommendations for future research in FPGA integration for IoT-based AI/ML applications were provided.