Edge Artificial Intelligence

22Design of a Deep Reinforcement Learning Approach for Optimization of Task Offloading and Resource Allocation for Edge Computing Networks

Anindita Khade¹* and Avaneesh Karthikeyan Iyer²

¹School of Technology Management and Engineering, SVKM’s NMIMS Deemed to be University, Navi Mumbai, Maharashtra, India

²Department of Computer Engineering, SIES Graduate School of Technology, Nerul, Navi Mumbai, Maharashtra, India

Abstract

Edge-cloud computing is an innovative framework that integrates edge and cloud computing capabilities to handle resource-intensive and time-critical tasks from mobile and IoT devices. Optimizing task offloading and resource allocation in these dynamic environments remains a complex challenge. This research presents a novel approach to this multi-objective optimization problem by employing a Double Deep Q-Network (DDQN) algorithm within the framework of Deep Reinforcement Learning (DRL). Our proposed DDQNEC (Double Deep Q-Network for Edge-Cloud) scheme models the task offloading and resource allocation issue is represented as a Markov Decision Process (MDP). By interacting with the edge-cloud ecosystem, the DDQN agent develops optimal decision-making strategies, taking into account resource usage, task requirements, and network conditions. The optimization objectives include maximizing resource utilization, minimizing task rejection rates, and reducing overall system costs. DDQNEC employs two neural networks - a prediction network and a target network to ...