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.

Distributed Machine Learning for Multiuser Mobile Edge Computing Systems.

Energy-Efficient Joint Offloading and Wireless Resource Allocation Strategy in Multi-MEC Server Systems.

Non-orthogonal Multiple Access Assisted Mobile Edge Computing Via Device-to-Device Communications

Secure Mobile Edge Computing Networks in the Presence of Multiple Eavesdroppers

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

Contextual User-Centric Task Offloading for Mobile Edge Computing in Ultra-Dense Network

Hybrid Online-Offline Learning for Task Offloading in Mobile Edge Computing Systems

Multi-Layer Computation Offloading in Distributed Heterogeneous Mobile Edge Computing Networks

Context-Aware Multi-User Offloading in Mobile Edge Computing: a Federated Learning-Based Approach

Task offloading mechanism based on federated reinforcement learning in mobile edge computing

BiLSTM-based Federated Learning Computation Offloading and Resource Allocation Algorithm in MEC

Decentralized Edge Learning via Unreliable Device-to-Device Communications

Performance Optimization in Mobile-Edge Computing Via Deep Reinforcement Learning

Distributed Energy Saving for Heterogeneous Multi-layer Mobile Edge Computing

Optimal dynamic spectrum allocation-assisted latency minimization for multiuser mobile edge computing

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

Dynamic Offloading Strategy for Delay-Sensitive Task in Mobile-Edge Computing Networks

Dynamic Parallel Multi-Server Selection and Allocation in Collaborative Edge Computing

Deep Reinforcement Learning-based Joint Optimization of Delay and Privacy in Multiple-User MEC Systems

Sequential Offloading for Distributed DNN Computation in Multiuser MEC Systems