.png){kind=icon}
.png){kind=icon}
Return
HAT: Task Offloading and Resource Allocation in RIS-Assisted Collaborative Edge Computing
.png){kind=icon}
PRE
AI
DOI：10.1109/TNSE.2024.3432893.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The problem of joint offloading decisions, resource allocation, and Reconfigurable Intelligent Surface (RIS) beamforming matrices for RIS-Assisted Edge Computing is a challenging issue. In this paper, user tasks can be either executed locally, or offloaded to a collaborative device or edge server with the assistance of the RIS, where RIS elements are grouped and assigned to all users to enable parallel services. The objective is formulated as a mixed integer nonlinear programming (MINLP) problem, where collaborative offloading decisions, RIS beamforming matrices, transmission power allocation, and computation resource allocation are jointly optimized to minimize the energy consumption. To address this problem, we propose a discrete-continuous Hybrid Action adapted Twin Delayed Deep Deterministic policy gradient (TD3) algorithm based on Deep Reinforcement Learning, named HAT. HAT constructs a latent representation space for the original discrete-continuous hybrid actions, fully considering the relations among highly coupled hybrid optimization variables. Experimental results demonstrate that HAT achieves significant performance gains over existing work (e.g., MELO, DDPG, PADDPG) and other benchmark schemes.
Keywords:
Array signal processing
Collaboration
Reconfigurable intelligent surfaces
Performance gain
Benchmark testing
Hybrid power systems
Resource management
Collaborative edge computing
deep reinforcement learning
reconfigurable intelligent surface
resource allocation
task offloading
Journal
I
IF:
7.9
Papers:
2.5K
Citations:
10.0K