Edge Computing Advancements

The field of edge computing is moving towards leveraging deep reinforcement learning and federated learning to improve task offloading and reduce latency. Researchers are exploring new approaches to enable efficient computation offloading, such as using Twin Delayed DDPG algorithms and knowledge-guided attention-inspired learning. Additionally, there is a focus on developing customized solutions for edge devices, including large language models and transformer-based image captioning models, to balance latency requirements with energy consumption and model accuracy. Noteworthy papers in this area include:

  • A Novel Deep Reinforcement Learning Method for Computation Offloading in Multi-User Mobile Edge Computing with Decentralization, which introduces a new approach based on the Twin Delayed DDPG algorithm to overcome the weaknesses of conventional DDPG-based power control strategies.
  • FAuNO: Semi-Asynchronous Federated Reinforcement Learning Framework for Task Offloading in Edge Systems, which presents a buffered, asynchronous federated reinforcement-learning framework for decentralized task offloading in edge systems.
  • CLONE: Customizing LLMs for Efficient Latency-Aware Inference at the Edge, which presents an in-depth algorithm-hardware co-design to accelerate the inference process and save energy while maintaining high-generation quality.

Sources

A Novel Deep Reinforcement Learning Method for Computation Offloading in Multi-User Mobile Edge Computing with Decentralization

Maximizing the Promptness of Metaverse Systems using Edge Computing by Deep Reinforcement Learning

FAuNO: Semi-Asynchronous Federated Reinforcement Learning Framework for Task Offloading in Edge Systems

CLONE: Customizing LLMs for Efficient Latency-Aware Inference at the Edge

Analyzing Transformer Models and Knowledge Distillation Approaches for Image Captioning on Edge AI

Knowledge-Guided Attention-Inspired Learning for Task Offloading in Vehicle Edge Computing

Built with on top of