Deep Reinforcement Learning for Task Scheduling in Intelligent Building Edge Network

With the development of information technology, intelligent devices and applications in the intelligent building environment have increasingly high requirements such as arithmetic power and time delay. The traditional cloud-based computing framework cannot meet these requirements. Using cloud-side collaborative scheduling can reasonably solve this problem. Task scheduling is the core problem of cloud-side collaboration, in recent years, there are many scheduling methods, while deep learning and reinforcement learning have attracted the attention of researchers because of their advantages of no human intervention and self-learning. However, they still have disadvantages such as slow convergence speed and low scheduling success rate. Therefore, this paper proposes an improved task scheduling algorithm based on deep reinforcement learning. We first establish a task scheduling algorithm framework based on deep reinforcement learning and then improve it by combining the Double Deep Q-Network (DDQN) method with the empirical replay method to improve the scheduling success rate and accelerate the convergence speed. We use standard dataset to validate this algorithm, and the experimental results show that this method has advantages over traditional task scheduling algorithms in terms of improving the scheduling success rate, accelerating the convergence speed, and reducing the overhead of edge servers.