Large language model assisted scheduling of collaborative aerial-ground system for emergency reconnaissance

EASIER is an ambitious interdisciplinary project that aims to address aerial-ground collaborative emergency response scheduling by integrating cutting-edge artificial intelligence with advanced optimization methods. Emergency response scenarios, such as flood reconnaissance, wildfire monitoring, or search-and-rescue, require rapid, flexible, and reliable scheduling solutions. However, current optimization methods are either too rigid, too expert-dependent, or insufficiently robust to handle the uncertainties and dynamic nature of real operations.To address these challenges, EASIER builds on three methodological pillars: (i) a large language model-assisted multi-agent scheduling framework that lowers the expertise barrier by translating natural-language mission requirements into structured problem definitions and solver configurations; (ii) a multi-objective optimization solver that combines knowledge-driven heuristics and data-driven transfer learning to efficiently balance competing performance criteria; and (iii) a robust optimization solver hybridized with reinforcement learning to effectively manage uncertainties in UAV endurance. The two solvers and existing exact solvers will be integrated into the scheduling framework and validated on a high-fidelity simulation platform, ensuring both technical feasibility and operational relevance.The originality of EASIER lies in its seamless combination of generative AI, optimization theory, and reinforcement learning in the context of aerial-ground collaboration, making it the first systematic attempt to bridge human natural-language intent with operational scheduling algorithms. The project will not only advance scientific knowledge but also deliver practical open-access tools that enhance the flexibility, robustness, and accessibility of emergency response services across Europe, in line with the EU Drone Strategy 2.0.