Disaster-Aware Informative Path Planning in Emergency Response Scenarios
Xinya Cheng, Na Liu
AI summary
Problem
Traditional path planning methods struggle to balance exploration efficiency with information value in dynamic emergency scenarios, often missing critical disaster zones due to rigid coverage strategies or reliance on inaccurate prior models.
Approach
The method integrates a Siamese UNet damage classifier with an ensemble uncertainty estimator to build a multi-layer map, which guides a Selective Frontier Algorithm to score and select paths based on damage severity, model confidence, redundancy, and flight cost.
Key results
- Outperforms coverage, random, and MCTS baselines in information coverage and semantic target hit rate
- Introduces a dynamic multi-layer map fusing building damage classification with model uncertainty
- Formulates a task-oriented information value function balancing damage level, uncertainty, redundancy, and flight cost
- Validates improved weighted information coverage on the xView2 remote sensing dataset
Why it matters
Enables UAVs to rapidly acquire high-value disaster information under constrained resources, directly supporting faster and more effective emergency rescue decision-making.
Abstract
In emergency response scenarios, rapid acquisition of critical disaster information supports effective decision-making. Tra- ditional geometric coverage-based path planning often strug- gles to balance efficiency and information value. To address this, we propose a Disaster-Aware Informative Path Plan- ning (DAIPP) method, which integrates a Siamese UNet- based building damage recognition model and formulates a novel information value function that considers recognition results, model uncertainty, and flight cost. We design an im- proved Frontier-based path planning algorithm, named the Selective Frontier Algorithm (SFA), which enhances the se- lection of candidate points to achieve the prioritized explo- ration of critical regions. To validate its effectiveness, the proposed method is compared with coverage path planning, random planning, and Monte Carlo tree search (MCTS). Ex- periments on the xView2 dataset demonstrate that the pro- posed method outperforms baselines in terms of information coverage, semantic target hit rate, and weighted information coverage, providing strong support for efficient disaster per- ception in emergency response.