Quantum-inspired bi-level neuro-swarm optimization for UAV-based disaster recognition and response

Reliable and timely situational awareness is essential for UAV-based disaster response yet learning-driven systems often face challenges arising from non-stationary environments, multimodal uncertainty, and limited onboard resources. The paper introduces the Quantum-Inspired Bi-Level Neuro-Swarm Optimization (QI-ENSO) method for unmanned aerial vehicle disaster relief in non-stationary and uncertain environments. The response to the calamity is treated as a bi-level problem that connects the uncertainties in perception with the decisions regarding the mission. QI-ENSO applies the learnable entanglement-based coordination and superposition-inspired exploration to reach stable convergence in the optimization landscapes characterized by non-convexity. The only algorithmic contribution made in this study is the QI-ENSO proposed method. All other aspects, like HDCA++ (the hierarchical dual-stage attention perception backbone), Digital Twin, reinforcement learning, and explainability tools, are just the non-modified and non-optimized modules that are used for evaluation and validation purposes only. Experiments conducted on multiple public UAV disaster datasets show that QI-ENSO achieves 3–7% improvements in F1-score, faster convergence, and low edge-level inference latency (≈3.2 ms) compared to recent adaptive swarm and hybrid learning baselines. The results show that quantum-inspired coordination mechanisms provide an effective and practical optimization strategy for UAV-based disaster response under uncertainty.In this research, the latency values were measured exclusively using desktop-class GPU hardware in a controlled laboratory setting. GPU-based execution on desktop-class hardware produced all the reported inference latency results; the direct deployment on embedded or onboard UAV platforms is not included in the present study and is mentioned as future work.