NextGen Resilient Network: Adaptive Simulation for Disaster Recovery

Abstract

When a major disaster unfolds, communication infrastructure is typically the first casualty and the slowest to be restored. This paper introduces a MANET-based framework engineered to sustain connectivity even when every fixed network element has been destroyed. The system was developed and validated in NS-3.45, employing a three-tier heterogeneous node structure: five stationary Network Towers that serve as primary Cluster Heads under normal conditions, fifteen slow-moving LoRa relay devices that assume the Cluster Head role once the towers are rendered non-functional, and 92 mobile user devices whose movement patterns draw directly from real KAIST pedestrian traces rather than any artificially generated mobility model. Dynamic cluster formation is driven by DBSCAN, which is re-invoked at every epoch, while AODV governs routing and path restoration. To probe self-healing behaviour, all five towers were brought down simultaneously at t?=?30?s. Within two seconds, the LoRa nodes had reorganised into a functional cluster topology and re-established connectivity for all 92 users, with no human intervention or pre-designated backup paths. Across the complete 100-second simulation, FlowMonitor recorded an average PDR of 88.32?%, throughput of 26.05?Kbps, and end-to-end delay of 0.104?s. These figures hold up favourably against a flat AODV baseline that was never exposed to complete infrastructure collapse.

Introduction

The text presents the design and motivation of a disaster-resilient Mobile Ad Hoc Network (MANET) system for emergency communications. Traditional communication infrastructure often fails during disasters such as earthquakes and floods, delaying rescue efforts. While MANETs avoid dependence on fixed infrastructure, they face challenges including dynamic topology, varying node density, and unrealistic assumptions in many previous studies that rely on synthetic mobility models and homogeneous nodes.

To address these limitations, the proposed system introduces three key improvements:

• Heterogeneous node roles, distinguishing towers, LoRa relays, and end-user devices.
• Real-world mobility, using KAIST pedestrian traces instead of random waypoint models.
• Complete infrastructure failure testing, measuring how quickly all 92 user nodes regain connectivity after a total network collapse.

The implementation is developed in NS-3.45 using AODV for routing, YansWifi for wireless communication, FlowMonitor for performance monitoring, and NetAnim for visualization.

The literature review identifies four major themes:

1. Fault-tolerant architectures improve network resilience through dynamic routing and decentralized control.
2. Mobility-aware clustering outperforms static approaches by adapting to changing node movement.
3. Real-world mobility traces provide more realistic and reliable simulation results than synthetic models.
4. Research gaps remain in heterogeneous node modeling, realistic mobility, and evaluation of recovery after total infrastructure failure.

The system architecture follows a structured UML-based design methodology, including requirements analysis, node role definition, validation, and refinement before coding. It consists of use case, context, and state diagrams that define system interactions and workflow. The high-level architecture integrates KAIST mobility traces, Gaussian node distribution, DBSCAN clustering, and NS-3 simulation, producing both visual outputs through NetAnim and quantitative performance metrics such as packet delivery ratio (PDR), delay, energy consumption, and recovery time.

Conclusion

The central motivation for this work was grounded in an observation about existing research: most MANET proposals for disaster recovery are evaluated under partial failure conditions, never under complete infrastructure collapse. The aim here was to construct a system capable of surviving the worst-case scenario and to measure exactly how it performs when pushed to that limit. Four objectives structured the work. The first established an architectural foundation through UML design artefacts. The second involved integrating DBSCAN-based dynamic clustering into NS-3, a task that proved more technically demanding than anticipated, particularly in ensuring stable re-clustering across the full simulation duration. The third produced a flat AODV baseline as a quantitative reference. The fourth was the definitive test: take down the entire infrastructure simultaneously and measure what the network is able to sustain. The most significant result was this: two seconds after all five towers were destroyed simultaneously, every one of the 92 user nodes had regained connectivity. Zero nodes remained isolated. The recovery was entirely self-directed: DBSCAN identified the tower failures, elected LoRa nodes as replacement Cluster Heads, and AODV reconstructed routing paths through the new topology without any external input or pre-configured fallback. The aggregate FlowMonitor results across the full 100-second run?- 88.32?% PDR, 26.05?Kbps throughput, and 0.104?s delay?- hold up well considering that the disaster interval is already factored into those figures. Flat AODV, which faced only partial random node failure and carried none of the clustering overhead, still produced lower PDR and substantially degraded throughput at scale. Several directions for future work suggest themselves naturally from this study. The most immediate is incorporating NS-3’s WifiRadioEnergyModel to quantify energy consumption and understand the lifetime implications of LoRa nodes bearing a permanent Cluster Head load. Beyond that, introducing UAV relay nodes would extend coverage into regions where ground-level node density is insufficient for DBSCAN to form stable clusters. Looking further ahead, reinforcement learning offers a promising path for Cluster Head election: rather than using degree centrality as a fixed heuristic, an RL-trained agent could learn to trade off delivery ratio, latency, and energy expenditure in proportion to whatever failure pattern the network is experiencing at any given moment.

References

[1] A. Verma et al., “Survey on disaster management communication networks,” IEEE Commun. Surveys & Tutorials, 2020. [2] S. K. Sharma and X. Wang, “Toward massive machine-type communications in disaster scenarios,” IEEE Access, 2019. [3] M. Conti et al., “Mobile ad hoc networking for emergency response,” IEEE Commun. Magazine, 2018. [4] H. Nguyen et al., “Resilient communication architectures for disaster recovery,” IEEE Network, 2021. [5] P. Kumar and R. Singh, “Mobility-aware routing protocols for disaster response networks,” IEEE Int. Conf. Advanced Networks, 2022. [6] L. Zhang et al., “Failure-tolerant network design for emergency communication systems,” IEEE Trans. Network & Service Mgmt., 2023. [7] R. Alotaibi et al., “Simulation-based analysis of disaster communication networks,” IEEE Access, 2024. [8] J. Lee and M. Park, “Adaptive network architectures for emergency and disaster management,” IEEE Systems Journal, 2025. [9] F. Khan et al., “Delay-tolerant networking for post-disaster communication: A survey,” IEEE Commun. Surveys & Tutorials, 2019. [10] T. D. Nguyen and M. Passarella, “Mobility models for disaster response scenarios in MANETs,” IEEE Systems Journal, 2021. [11] A. Al-Fuqaha et al., “Machine learning for adaptive network routing in disaster environments,” IEEE Network, 2022. [12] S. Hamida et al., “Cross-layer design for resilient communication in emergency MANETs,” IEEE Trans. Mobile Computing, 2023. [13] R. Li and Y. Liu, “Simulation-based performance evaluation of ad hoc disaster networks,” IEEE Access, 2024. [14] J. A. Khan et al., “Mobility-aware clustering protocols for emergency communication systems,” IEEE ICC, 2024. [15] H. Lee et al., “Disaster resilience in heterogeneous wireless networks with AI-assisted control,” IEEE Trans. Wireless Commun., 2025. [16] NS-3 Network Simulator Documentation. [Online]. Available: www.nsnam.org, 2024. [17] CRAWDAD Dataset Repository. [Online]. Available: crawdad.org. [18] J. Zhang et al., “A resilient routing strategy based on deep reinforcement learning for MANETs,” Computer Networks, 2024. [19] T. Matsuda et al., “Scalable routing in city-scale Wi-Fi for disaster recovery,” arXiv, 2025. [20] S. Caleb et al., “Enhancing fault tolerance in wireless mesh networks for disaster management,” IEEE, 2023.

Copyright

Copyright © 2026 Prof. Ruchitha K J, Monisha D R, Shaik Saaduddin, Srushti Waghamode, Darshan B Gowda. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.