Reinforcement Learning

This paper explores energy-efficient UAV swarm-assisted mobile edge computing (MEC) with dynamic clustering and scheduling. UAVs are divided into swarms, each with a leader and followers, serving ground users. Unlike previous works, UAVs can change swarms dynamically based on spatial positions and application placements. We aim to maximize long-term energy efficiency by optimizing UAV clustering and scheduling, reformulating the problem as interconnected multi-agent stochastic games and proving a Nash Equilibrium (NE). We propose a reinforcement learning-based algorithm (RLDC) for coordination, analyzing its convergence and complexity. Simulations show RLDC’s superior performance over existing methods.

This paper introduces a new reputation-enhanced shard-based Byzantine fault-tolerant scheme tailored for zero-trust human digital twin (HDT) systems. This distinctive framework seeks to improve throughput and diminish latency in data sharing by integrating a parallel validation-based reputation-enhanced practical Byzantine fault tolerance consensus process with a priority-based block-adding method to counter forking attacks. Overall, the approach uses a Q-network to handle the considerable dimensions of the action space in the problem at hand. Consequently, this solution effectively enriches the authentication, authorization, and verification endeavors in HDT systems via an increase in throughput and a decrease in latency, presenting a potent answer to secure and effectual data interchange in HDT setups.

This paper introduces a secure and efficient connectivity scheme for human digital twin (HDT) systems. It combines differential privacy, federated multi-task learning, and blockchain to ensure accuracy, privacy, and reduced connectivity costs. The proposed scheme enhances security and compares favorably to other solutions for HDT networks.

This paper focuses on energy-efficient scheduling for multiple UAV-assisted MEC. It optimizes UAV trajectories, energy renewal, and application placement to maximize long-term energy efficiency. The approach uses a triple learner based reinforcement learning method and demonstrates superiority through simulations.

This paper focuses on an energy efficient scheduling problem for multiple unmanned aerial vehicles (UAVs) that assist in mobile edge computing. The goal is to maximize the long-term energy efficiency of the UAVs by optimizing their trajectory planning, energy renewal, and application placement. The paper proposes a triple learner based reinforcement learning approach to address the problem, which includes a trajectory learner, an energy learner, and an application learner. Simulations show that the proposed solution outperforms existing approaches.

This paper addresses the problem of joint sensor activation and mobile charging vehicle scheduling for wireless rechargeable sensor networks in industrial Internet of Things (IIoT). The goal is to optimize the system energy consumption while meeting quality-of-monitoring requirements and sensor charging deadlines. The paper proposes a novel scheme that combines reinforcement learning and approximation algorithms to solve the problem efficiently. Simulation results demonstrate the feasibility and superiority of the proposed scheme over existing approaches.

This paper focuses on the joint optimization of sensor activation and mobile charging scheduling in industrial wireless rechargeable sensor networks (IWRSNs). The goal is to minimize energy consumption while meeting task requirements, sensor charging deadlines, and the mobile charger vehicle’s energy capacity. The proposed solution combines deep reinforcement learning and a marginal product-based approximation algorithm. Simulation results show the superiority of this solution compared to other methods.