Enhancing Disaster Resilience with UAV-Assisted Edge Computing: A Reinforcement Learning Approach to Managing Heterogeneous Edge Devices DOI: XXXXXXX.XXXXXXX Journal: JATS Volume: 37 4 111 11 CCS: Applied computing Transportation CCS: Computing methodologies Reinforcement learning
Talha Azfar, Kaicong Huang, Ruimin Ke · 2025
Read Enhancing Disaster Resilience with UAV-Assisted Edge Computing: A Reinforcement Learning Approach to Managing Heterogeneous Edge Devices DOI: XXXXXXX.XXXXXXX Journal: JATS Volume: 37 4 111 11 CCS: Applied computing Transportation CCS: Computing methodologies Reinforcement learning by Talha Azfar, Kaicong Huang, Ruimin Ke (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-15305.txt.
Introduction
Talha Azfar, Kaicong Huang, and Ruimin Ke use drones as flying edge computers to keep disaster sensor networks alive longer. Quakes, floods, and fires knock out power and base stations just when field data matters most. A drone can lend its battery and its radio to the stranded nodes.
Their 2025 paper trains a reinforcement learning agent to juggle drones, power, and weak links across many failure scenes. The agent learns which device is most likely to die next, a ranking that tells repair crews where to go first. Two staged evacuations, a rural town and a city downtown, test the plan. Network life stretched where it counted most, the authors report.
Catalog record
| CREATOR | Talha Azfar, Kaicong Huang, Ruimin Ke |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-15305
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Enhancing Disaster Resilience with UAV-Assisted Edge Computing: A Reinforcement Learning Approach to Managing Heterogeneous Edge Devices DOI: XXXXXXX.XXXXXXX Journal: JATS Volume: 37 4 111 11 CCS: Applied computing Transportation CCS: Computing methodologies Reinforcement learning Talha Azfar, Kaicong Huang, Ruimin Ke Abstract . Edge sensing and computing is rapidly becoming part of intelligent infrastructure architecture leading to operational reliance on such systems in disaster or emergency situations. In such scenarios there is a high chance of power supply failure due to power grid issues, and communication system issues due to base stations losing power or being damaged by the elements, e.g., flooding, wildfires etc. Mobile edge computing in the form of unmanned aerial vehicles (UAVs) has been proposed to provide computation offloading from these devices to conserve their battery, while the use of UAVs as relay network nodes has also been investigated previously. This paper considers the use of UAVs with further constraints on power and connectivity to prolong the life of the network while also ensuring that the data is received from the edge nodes in a timely manner. Reinforcement learning is used to investigate numerous scenarios of various levels of power and communication failure. This approach is able to identify the device most likely to fail in a given scenario, thus providing priority guidance for maintenance personnel. The evacuations of a rural town and urban downtown area are also simulated to demonstrate the effectiveness of the approach at extending the life of the most critical edge devices. 1. Introduction In the face of natural disasters, remote locations, or emergency scenarios, the availability of reliable infrastructure remains a critical challenge. Traditional infrastructure systems often falter in such situations due to power outages, communication breakdowns, and the inability to conduct timely maintenance. In these contexts, edge infrastructure devices play a pivotal role in maintaining connectivity, collecting vital data, and facilitating emergency response efforts in urban areas ( Higashino et al. 2017 ) and remote hostile environments ( Hussain et al. 2019 ) . Edge computing devices usually perform crucial tasks involving sensing, processing, and communication while consuming relatively low power. This could be as simple as recording and updating the temperature, or as complex as real-time multi-vehicle detection and tracking from surveillance cameras. These edge devices can be instrumental in disaster response by monitoring the situation, detecting urgent areas of concern, and coordinating the emergency response ( Aboualola et al. 2023 ) . Some challenges faced by these edge devices in remote locations are sporadic or otherwise limited power supply and communication channels. Similarly, in disaster scenarios such as floods or wildfires, power grids and wireless communication base stations may partly or completely fail, leaving the edge devices to run on backup power and wait for alternative communication channels such as mobile base stations until the usual systems are restored. However, getting to the locations of the devices and providing the necessary services may not be feasible in the short term, especially in emergency situations where other tasks may take priority. Recent works have proposed an innovative solution that tackles both the power consumption and communication constraints in the form of an unmanned aerial vehicle (UAV) mobile edge computing system ( Wang et al. 2021 ) , in which a UAV flies within communication range of the edge devices and offloads their computation tasks thus saving their battery, and simultaneously acts as a high gain relay node to the base station which is not reachable by the low power edge devices. The UAV is not bound by the same limitations as the devices in the field since it can return to its base for charging at any time. While UAV relay nodes as part of the communication network and UAV mobile edge computing nodes have been considered separately, the idea of data criticality in emergency situations needs more attention. During a disaster recovery process, intelligent edge infrastructure may provide crucial data to responders, and is only useful if the communication is up to date. Data criticality then emerges as a pivotal concern in these networks, reflecting the diverse and important nature of data handled. If power and communications are both affected by the disaster, then there is no guarantee that data is being received from all edge nodes on time. This motivates the construction of a strategy that pays special attention to the age of the uncommunicated data held in the edge devices that have lost communication. Together with the power loss constraint, it becomes an interesting problem to study with reinforcement learning to discover the best strategy in any given configuration of devices. Another important aspect missing from the existing literature is the simulation of the direct impacts of natural disasters and the associated evacuation of individuals utilizing road networks. Furthermore, it raises the critical question of how to prioritize edge devices based on their utility and effectiveness in such scenarios. In this paper we explore the feasibility and efficacy of integrating UAVs and reinforcement learning (RL) techniques into the deployment and management of edge infrastructure devices in rural areas and emergency scenarios by varying the possible constraints and configurations to reflect real-world challenges. Specifically, we address the issue of losing power and communication in disaster stricken areas and show that the UAV can learn to extend the life of the devices. First, the problem is mathematically formulated with physical constraints on the device energy and data age. Then, we use reinforcement learning with different reward functions to identify the most general method of solving the problem. We also simulate the evacuation of a rural town and an urban downtown area, and study how the UAV can prioritize edge devices based on their proximity to the traffic flow. The contributions of this paper are summarized as follows: • Handling Heterogeneous Edge Devices: The paper addresses the challenge of managing a diverse set of edge devices with varying computational capabilities, power constraints, and communication abilities proposing a UAV-assisted system to uniformly manage these differences while maintaining network resilience. • Mitigating Power and Communication Outages: By utilizing UAVs to provide both computational offloading and communication relays, the study offers a solution to the frequent power and connection outages experienced by edge devices during disasters, ensuring continuous data collection and connectivity. • Reinforcement Learning for Strategic Optimization: The system employs deep Q-network reinforcement learning to develop strategies for optimizing UAV operations, including prioritizing critical data delivery and extending the operational life of the network in various disaster scenarios. • Scenario-Based Evaluation: The study conducts extensive simulations, including realistic disaster evacuation scenarios, to demonstrate the effectiveness of the proposed approach in prioritizing critical data delivery and maintaining infrastructure functionality under diverse conditions of power and communication availability. 2. Literature Review Researchers have long explored the integration of UAVs with cellular network infrastructure to optimize energy efficiency while ensuring the maintenance of quality of service ( Naqvi et al. 2018 ) ( Tran et al. 2022 ) , while the mixed architecture of terrestrial and aerial connectivity has featured in numerous studies about infrastructure resilience in the face of disasters for preparedness, assessment, and response and recovery ( Erdelj et al. 2017 ) . UAVs have been considered as relay nodes in communication networks to