Alcides Fernandes

Alcides Fernandes

Auto-generated profile, based on publicly available information

Alcides Fernandes


Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

8 May 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
AFAlcides Fernandes
Alcides Fernandes
Hugo LandaluceHugo Landaluce
Hugo Landaluce
IAIgnacio Angulo
Ignacio Angulo
JMJon Munoz
Jon Munoz
Maximizing the performance of real-time AI inference at the edge is essential for enabling intelligent industrial systems. While conventional strategies rely on AI model optimization and hardware acceleration, performance in edge AI architectures can also be improved by reducing communication latency, minimizing protocol overhead, optimizing CPU and energy usage, and ensuring high-throughput data exchange between field devices and embedded platforms.This work investigates the implementation constraints of IIoT protocols on edge devices and analyzes communication protocol selection as a critical factor in achieving high-performance inference. A comparative evaluation is conducted on widely used protocols including MODBUS, OPC-UA, MQTT, HTTP, CoAP, TCP, and UDP, deployed on Siemens PLCs and an NVIDIA Jetson Xavier within a Gigabit Ethernet infrastructure.Latency, throughput, CPU utilization, and power consumption are measured under varying payload sizes and real-time conditions. The results demonstrate that protocol choice directly impacts inference efficiency, and that hardware limitations such as restricted payload handling, limited CPU resources, and fragmentation overhead in PLCs which can significantly constrain performance. TCP and UDP deliver the best results for time-sensitive applications, while MQTT and OPC-UA provide trade-offs between interoperability and resource usage.This study offers a practical framework for protocol selection that aligns communication strategies with hardware capabilities and AI workload demands in edge computing environments.