Publication

Abstract : The difficulty with dynamic and heterogeneous natured edge computing environments is resource provisioning. Reinforcement Learning (RL) can be promising to solve the problems of resource allocation under conditions of dynamics and complexity. This research work presents a novel DDPG-based (Deep Deterministic Policy Gradient) approach for optimizing the dynamics of resource allocation in edge computing environments by exploiting the CRAWDAD KAIST–WIBRO dataset for the proposed model. This approach implements a neural network that learns the optimal policies of allocations with dynamism towards new changes in workloads and resource availabilities. In order to test the performance of this method, conducted several simulations over real-world workload and resource data, mapping the parameters like rewards and variation in episodes. The simulation results indicate that proposed approach performs superior compared to traditional heuristic-based approaches as well as other RL algorithms for resource usage, response time, and energy efficiency. This proposed approach permits optimized resource utilization along with quality of service, of paramount importance in executing complex and demanding applications at the edge computing environment.