Aiming in Harsh Environments: A New Framework for Flexible and Adaptive Resource Management

TL;DR

This paper introduces a novel adaptive resource management framework for harsh environments, utilizing deep learning for environment prediction and self-organization for service management, demonstrated through a case study.

Contribution

It presents a new network architecture with environment-aware modules, including deep learning-based prediction and self-organized management, addressing challenges in harsh, unattended environments.

Findings

Effective environment resource prediction using deep learning

Optimized resource allocation based on environmental predictions

Demonstrated improvements in system utility in case study

Abstract

The harsh environment imposes a unique set of challenges on networking strategies. In such circumstances, the environmental impact on network resources and long-time unattended maintenance has not been well investigated yet. To address these challenges, we propose a flexible and adaptive resource management framework that incorporates the environment awareness functionality. In particular, we propose a new network architecture and introduce the new functionalities against the traditional network components. The novelties of the proposed architecture include a deep-learning-based environment resource prediction module and a self-organized service management module. Specifically, the available network resource under various environmental conditions is predicted by using the prediction module. Then based on the prediction, an environment-oriented resource allocation method is developed to optimize the system utility. To demonstrate the effectiveness and efficiency of the proposed new functionalities, we examine the method via an experiment in a case study. Finally, we introduce several promising directions of resource management in harsh environments that can be extended from this paper.