Green Orchestration of Digital Twin Analytics: Joint Optimization of AI, IoT, and 5G/6G Resources

Abstract

Digital Twin (DT) systems built on the Internet of Things (IoT), artificial intelligence (AI), and 5G/6G networks continuously mirror physical processes, but the energy and carbon cost of keeping them synchronized is rarely optimized end to end. This paper proposes a Green Orchestration framework, the Green Adaptive Multi-Objective Digital Twin Resource Optimization (GAMDTRO) architecture, which jointly optimizes AI model complexity, edge-cloud placement, IoT resource allocation, 5G/6G communication resources, DT synchronization interval, and carbon-aware scheduling in one multi-objective formulation. GAMDTRO couples a Non-dominated Sorting Genetic Algorithm II (NSGA-II) outer loop for structural decisions with a Multi-Agent Proximal Policy Optimization (MAPPO) inner loop for real-time power and bandwidth control, gated by grid carbon-intensity signals. Fifteen equations model energy, carbon, latency, synchronization staleness, accuracy-complexity trade-off, quality of service (QoS), cost, and the multi-objective function, with associated constraints. Simulation-style evaluation across 100 to 1000 IoT devices shows energy, carbon, and cost reductions of up to 34%, 41%, and 38% against a cloud-only baseline, with competitive latency and over 95% of achievable accuracy retained.