Abstract:This paper studies a sequential task offloading problem for a multiuser mobile edge computing (MEC) system. While most of the existing works consider static one-shot offloading optimization with fixed wireless channel conditions and fixed computational tasks, we consider a dynamic optimization approach, which embraces wireless channel fluctuations and random deep neural network (DNN) task arrivals over an infinite horizon. Specifically, we introduce a local CPU workload queue (WD-QSI) and an MEC server workload queue (MEC-QSI) to model the dynamic workload of DNN tasks at each WD and the MEC server, respectively. The transmit power and the partitioning of the local DNN task at each WD are dynamically determined based on the instantaneous channel conditions (to capture the transmission opportunities) and the instantaneous WD-QSI and MEC-QSI (to capture the dynamic urgency of the tasks) to minimize the average latency of the DNN tasks. The joint optimization can be formulated as an ergodic Markov decision process (MDP), in which the optimality condition is characterized by a centralized Bellman equation. However, the brute force solution of the MDP is not viable due to the curse of dimensionality as well as the requirement for knowledge of the global state information. To overcome these issues, we first decompose the MDP into multiple lower dimensional sub-MDPs, each of which can be associated with a WD or the MEC server. Next, we further develop a parametric online Q-learning algorithm, so that each sub-MDP is solved locally at its associated WD or the MEC server. The proposed solution is completely decentralized in the sense that the transmit power for sequential offloading and the DNN task partitioning can be determined based on the local channel state information (CSI) and the local WD-QSI at the WD only. Additionally, no prior knowledge of the distribution of the DNN task arrivals or the channel statistics will be needed for the MEC server. The proposed solution can achieve the superb performance over various state-of-the-art baselines.

Dynamic computation offloading in time-varying environment for ultra-dense networks:a stochastic game approach

Delay-Optimal Computation Offloading for Computation-Constrained Mobile Edge Networks.

Delay-Optimal Computation Offloading in Large-Scale Multi-Access Edge Computing Using Mean Field Game

Dynamic Computation Offloading for Mobile Cloud Computing: A Stochastic Game-Theoretic Approach

Dynamic Offloading for Multiuser Muti-CAP MEC Networks: A Deep Reinforcement Learning Approach

Optimized Computation Offloading Performance in Virtual Edge Computing Systems via Deep Reinforcement Learning

Dynamic Computation Offloading and Server Deployment for UAV-Enabled Multi-Access Edge Computing

Joint Task Offloading, D2D Pairing, and Resource Allocation in Device-Enhanced MEC: A Potential Game Approach

Resource Allocation for NOMA-MEC Systems in Ultra-Dense Networks: A Learning Aided Mean-Field Game Approach

Multi-User Computation Offloading in Mobile Edge Computing: A Behavioral Perspective

Stochastic Congestion Game for Load Balancing in Mobile-Edge Computing

AI-Enabled Task Offloading for Improving Quality of Computational Experience in Ultra Dense Networks

Distributed Task Offloading Game in Multiserver Mobile Edge Computing Networks

Joint Game Theory and Greedy Optimization Scheme of Computation Offloading for UAV-Aided Network

Online Computation Offloading and Resource Scheduling in Mobile-Edge Computing

NOMA-Based Multi-User Mobile Edge Computation Offloading via Cooperative Multi-Agent Deep Reinforcement Learning

Sequential Offloading for Distributed DNN Computation in Multiuser MEC Systems

Efficient Multi-User Computation Offloading for Mobile-Edge Cloud Computing

A Hybrid DQN and Optimization Approach for Strategy and Resource Allocation in MEC Networks

Decentralized Computation Offloading for Multi-User Mobile Edge Computing: A Deep Reinforcement Learning Approach

Dynamic Computing Offloading Strategy in LEO Constellation Edge Computing Network