Quick and Good: A DRL Based Communication-Caching-Energy Joint Optimization Scheme for Prolonging the Lifetime of UAV Assisted IoE

Abstract

The rapid increase in the number of Internet of things (IoT) devices has led to significant access pressure, making network energy consumption and communication load key challenges. Edge caching, cooperative communication, and energy management technologies have proven to be effective in alleviating these issues. This paper investigates a unmanned aerial vehicle (UAV)-assisted Internet of everything (IoE) architecture that integrates caching, communication, and energy management. A collaborative communication-caching-energy optimization scheme is proposed, which involves the joint operation of the UAV and base station (BS) to pre-cache content required by ground users, thus minimizing system energy consumption. We model the joint optimization of content caching, communication, and energy consumption as a Markov decision process (MDP), transforming it into a long-term optimization problem solvable by deep reinforcement learning. Based on the simple deep Q-network (DQN), we design a dynamic content placement strategy that jointly optimizes communication, caching, and energy consumption. Simulation results demonstrate that the proposed method, compared to branch and bound (B&B), particle swarm optimization (PSO), genetic algorithm (GA), and random algorithms, not only approaches the optimal solution most closely, effectively reducing system energy consumption, but also exhibits the lowest time complexity.