Deep Q-Network Based Task Scheduling Strategy in Computing First Network

In order to solve the problem of server overload due to uneven resource allocation and total resource constraint, there is a pressing need to study resource allocation and task scheduling strategies in CFN (computing first network). A scheduling algorithm based on PSTDQN (Priority Sort Tasks DQN) is proposed to solve this problem. Firstly, a model for task scheduling decision and arithmetic resource demand allocation is constructed, which aims to optimize the resource allocation and routing control in the task scheduling process; Secondly, Taking the minimum delay of the task as the optimization objective, the convergence rate and searchability of the DQN algorithm are improved by introducing task priority and variability strategies. Finally, Solve the optimal scheduling scheme using the improved PSTDQN algorithm. compared to D3QN and DDPG, PSTDQN performs more optimally in key metrics such as convergence speed, task completion delay, resource utilization rate and task completion, confirming its validity in task scheduling in CFN.