Green AI for Edge and Distributed Intelligent Systems

Artificial intelligence now runs far beyond the data centre: on sensors, embedded controllers, vehicles, industrial equipment, mobile devices and networks of cooperating edge nodes. These systems must learn and act under strict limits on energy, memory, bandwidth and compute, yet the dominant development practice still rewards predictive accuracy while treating energy use, carbon footprint and hardware demand as afterthoughts. Green AI proposes a different balance, in which accuracy, energy, latency, communication cost and environmental impact are optimised together. Recent progress in model compression, adaptive inference, TinyML, nature-inspired optimisation and communication-efficient federated learning shows that this balance is achievable. What is missing is a coherent body of work that evaluates these techniques against explicit energy budgets on real, constrained hardware, using measurements that can be compared across platforms, workloads and deployment settings.This Research Topic aims to establish energy and resource consumption as primary design objectives for machine learning on resource-constrained and distributed intelligent systems. It will bring together methods that shrink, adapt, schedule and coordinate learning models so they deliver reliable performance within the power, memory and communication limits of edge, IoT and cyber-physical deployments. We are especially interested in optimisation-driven approaches—including metaheuristic, evolutionary and learning-based control strategies—that jointly optimise accuracy, energy, latency and carbon cost; in federated and distributed learning that remains efficient and secure when clients are heterogeneous and energy-limited; and in reproducible ways to measure and benchmark the sustainability of intelligent systems. Every contribution should report measured energy consumption or resource-efficiency outcomes as a central result, based on realistic workloads, and should make a clear contribution to machine-learning or intelligent-system design rather than to infrastructure engineering or policy alone.Topics of interest include, but are not limited to:• Energy-efficient learning algorithms and training strategies for constrained hardware• Model compression, quantisation, pruning and distillation for embedded and edge deployment• TinyML and on-device intelligence for IoT and cyber-physical systems• Metaheuristic, evolutionary and learning-based optimisation of models under energy budgets• Sustainable federated and distributed learning with limited energy and communication• Carbon-aware scheduling and resource allocation across the edge-cloud continuum• Hardware-aware optimisation for microcontrollers, accelerators and heterogeneous platforms• Benchmarks, measurement protocols and open tools for AI sustainability metrics• Efficiency, security and privacy trade-offs in distributed intelligent systems• Real-world Green AI deployments in industry, mobility, agriculture and environmental monitoring Topic editors: JAGANNATH NALAVADE, Rajesh Prasad, Ramachandra Pujeri, Youakim Badr Source: https://www.frontiersin.org/research-topics/85651/green-ai-for-edge-and-distributed-intelligent-systems