AI-driven control systems for embedded devices: Evolution and impact
Year: 2025
Type: article
Abstract: Artificial intelligence is fundamentally transforming power and performance control mechanisms in embedded systems across numerous industries. Traditional control methodologies like PID controllers and model-based approaches have long dominated embedded applications but face inherent limitations when managing complex, nonlinear system dynamics. The integration of neural networks and machine learning techniques represents a significant advancement that addresses these challenges. These AI-enhanced control systems demonstrate superior adaptation speed, accuracy, and disturbance rejection while capturing intricate relationships between system variables without explicit modeling requirements. Despite higher initial development investments, these intelligent controllers offer substantial long-term benefits through reduced maintenance needs and enhanced performance. Hybrid architectures combining conventional control theory with machine learning show particular promise by leveraging the predictability of traditional approaches alongside the adaptability of neural networks. As embedded processors continue advancing, on-device learning capabilities will enable unprecedented personalization and efficiency, with systems adapting to usage patterns, component aging, and environmental factors in real-time. The standardization of interfaces, pre-trained models, and optimization tools for resource-constrained environments will accelerate industry adoption, ultimately revolutionizing how embedded devices balance performance requirements with power constraints.
Author: Patel, Pratikkumar Dilipkumar
Institution: Arizona State University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7090841744
FWCI:
Cites:
Cited by:
Related to: