,
Assistant Professor, Kalinga University , Naya Raipur, Chhattisgarh , India
Assistant Professor, Kalinga University, , Naya Raipur, Chhattisgarh , India
Rapid and accurate assessment of civic infrastructure following a natural or artificial disaster is essential to planning emergency response and recovery. This paper introduces a control system based on deep reinforcement learning (DRL) to coordinate unmanned aerial vehicle (UAV) swarms and methodically approach the post-disaster infrastructure inspection. The multi-UAV coordination problem is formulated as a cooperative Markov decision process, enabling the learning of optimal policies for navigation, coverage, and collision avoidance under highly dynamic, uncertain conditions in disasters. The training-and-decentralized-execution paradigm is centralized to provide scalable swarm behavior while retaining real-time operational feasibility. The simulation experiments are conducted in real post-disaster urban settings marked by damaged structures, blocked streets, and limited communication. The average spatial coverage of the proposed DNR-controlled swarm is 91.6 decision steps, which is better than that of the rule-based and heuristic baselines (138.4 and 126.7 decision steps, respectively). The trained policy incurs a 34.2% lower cumulative navigation cost and maintains a stable inter-UAV separation, with a variance of less than 0.12 across multiple trials. Convergence of the policy is obtained in 2,150 training episodes, which is more than 3,900 training episodes in the case of baseline learning methods. The statistical analysis of 50 simulation runs indicates that dispersion in mission completion time was reduced by 27.5% and coverage uniformity improved by 22.8%. Moreover, the trained system shows robustness to partial failures of UAVs and adaptable obstacles, as training is not needed. These results verify that deep reinforcement learning offers a powerful and effective tool for autonomous swarm UAV deployment in post-disaster civil infrastructure inspection, aiding timely situational awareness and evidence-based decision-making within disaster management agencies.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.