Abstract
Task offloading could optimize computational resource utilization in edge computing environments. However, how to assign and offload tasks for different behavior users is an essential problem since the systems dynamic, intelligent application diversity, and user personality. With user behavior prediction, this paper proposes soCoM, a semi-online Computational Offloading Model. We explore the user behaviors in sophisticated action space by reinforcement learning for catching unknown environment information. With Dueling Deep-Q Network, both the prediction accuracy of users' behaviors and the server load balance are well-considered, while increasing the computational efficiency and decreasing the resource costing. We propose a dynamic simulation environment of edge computing to demonstrate that user behavior is the critical factor for impacting system utilization. As the action space increasing, Dueling DQN performs better than state-of-art DQN and other improved strategies, and also load balance in multiple different server scenario.
Highlights
The recent advancement of Internet of Things (IoT) has motivated various applications with different requirements (e.g. Automatic Speech Recognition (ASR), Nature Language Processing (NLP), Computer Vision (CV) and accurate human-computer interaction (HCI)) [1]–[3]
Computation offloading from edge users to servers becomes an essential part of mobile edge computing (MEC)
The main contributions of this paper are as follows: 1) We propose SEMI-ONLINE COMPUTATIONAL OFFLOADING MODEL (soCoM), which offloads tasks reasonably by predicting user equipment (UE) behavior differences without further communications by guidelines or labeled data. soCoM is a semi-online distributed offloading model for an edge computing system. soCoM utilizes Markov Decision Process and reinforcement learning to make decisions and explore the unknown dynamic user information
Summary
The recent advancement of Internet of Things (IoT) has motivated various applications with different requirements (e.g. Automatic Speech Recognition (ASR), Nature Language Processing (NLP), Computer Vision (CV) and accurate human-computer interaction (HCI)) [1]–[3]. S. Song et al.: Semi-Online Computational Offloading by Dueling DQN for User Behavior Prediction TABLE 1. Challenges in computation offloading procedure: i) the state space of the UE set would increase exponentially as the growth of UE number (shown in FIGURE 1); ii) the UE behavior observation is more difficult for server part since the diversity between edge users. To offload the tasks between different UE and servers with high performance, in this paper, we introduce semi-online architectures and Dueling Deep-Q Network (Dueling DQN) [13] to optimize resource utilization, energy consumption, and network latency in computation offloading procedures. Dueling Deep-Q Network could optimize two aims in ample action space: user behavior prediction and server load balance.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call