Environmental emergency scenarios often involve dynamic hazards whose progression is difficult to predict due to inherent uncertainties and time dependencies. Although mathematical models exist to approximate these dynamics, they are often incomplete or imprecise. Consequently, real-time data are crucial for refining understanding of the situation, enhancing situation awareness and improving response strategies. Recent advances in autonomous unmanned systems, such as drones or tracked robots, have enabled rapid, on-the-ground data collection. These systems often offer cost-effective and flexible means to continuously update models, allowing for adaptive monitoring that can track evolving hazards. However, the variety of systems, application domains, and methodological approaches spread across various research domains presents a challenge for future research. This work provides a perspective and structured synthesis of techniques for integrating in situ observations into adaptive models for situation awareness, framed through an overarching sequential decision-making view and a control hierarchy for autonomous unmanned systems. Specifically, it focuses on recurring design principles in common emergency application contexts such as wildfires, CBRN, and flood and storm systems, and in adaptive information gathering and data fusion methods. Common principles that have emerged largely independently across domains are highlighted and key challenges that originate in this heterogeneity of approaches are identified. Particular attention is given to their transferability to real-world applications. Therefore, this study organizes fragmented research fields, offering a unifying set of concepts, limitations, and future perspectives for adaptive monitoring of dynamic environmental hazards with unmanned systems in emergency scenarios.
Gioia et al. (Thu,) studied this question.