DDPG-Based Edge Resource Management for Coal Mine Surveillance Video Analysis in Cloud-Edge Cooperation Framework

Abstract

Intelligent video surveillance is important to ensure production safety in coal mines, while cloud-edge cooperation is an effective means to improve the performance of intelligent video monitoring. However, in edge layers, incorrect resource allocation of computing and network resources will result in the waste of resources and low real-time performance. In this paper, a DDPG-Based (Deep deterministic policy gradient-based) edge resource allocation method for cloud-edge cooperation framework is proposed. Firstly, the cloud-edge cooperation framework is designed for different tasks. Secondly, the joint minimizing problem of latency and bandwidth usage caused by edge computing is modeled. To quickly solve the joint optimization problem, we convert it to MDP (Markov Decision Process). In addition, ESPN (Edge status perception network) is proposed, which enhances the ability of feature perception and action output of DDPG. Finally, DDPG-ESPN is proposed to solve the joint optimization problem. Simulation results show that compared with other methods, DDPG-ESPN improves the real-time performance and bandwidth usage by up to 18.88% and 42.81% respectively.

Keywords

• Edge resource allocation
• intelligent video surveillance
• deep deterministic policy gradient
• edge status perception network