Artificial intelligence for energy-efficient computation offloading in WPT-enabled industrial internet of things

Area of Science:

• * Computer Science
• * Electrical Engineering
• * Artificial Intelligence

Background:

• * Resource-constrained Internet of Things (IoT) systems require intelligent edge computing and Wireless Power Transfer (WPT) for real-time decision-making and sustainable Industrial Internet of Things (IIoT) operations.
• * AI-driven WPT enhances time-division multiplexing (TDM) efficiency for data offloading, improving system sustainability and efficiency with intelligent edge computing.
• * Stringent battery capacity constraints in WPT systems necessitate novel approaches to relax energy limitations.

Purpose of the Study:

• * To propose a perturbation-based virtual energy queue to overcome battery capacity constraints in WPT systems.
• * To reformulate the Energy Efficiency (EE) optimization problem into a drift-plus-penalty framework using Dinkelbach's transformation for reduced latency and stable queues.
• * To introduce a hybrid Deep Reinforcement Learning (DRL)-Lyapunov optimization framework for adaptive online scheduling, power minimization, and stable queue management.

Main Methods:

• * Implementation of a perturbation-based virtual energy queue to eliminate the need for future system condition prediction.
• * Application of Dinkelbach's transformation to convert the long-term EE optimization into a drift-plus-penalty framework.
• * Development of a hybrid DRL-Lyapunov optimization framework with an actor-critic model for dynamic Central Processing Unit (CPU) frequency adjustment and optimal policy learning.

Main Results:

• * The proposed hybrid DRL-Lyapunov framework demonstrated a 15-20% enhancement in Energy Efficiency (EE) compared to a fixed-power baseline.
• * Device battery levels were maintained within an optimal range of 55-60%, ensuring energy balance.
• * The framework effectively ensured queue stability and reduced variations in battery dynamics.

Conclusions:

• * The hybrid DRL-Lyapunov framework offers a sustainable solution for Industrial Internet of Things (IIoT) operations by optimizing energy efficiency and battery management.
• * The proposed approach effectively addresses battery capacity constraints and latency requirements in resource-constrained WPT systems.
• * This AI-driven framework enables intelligent, adaptive, and stable online scheduling for enhanced system performance and longevity.