Priority-Aware Multi-UAV Search Path Planning with Evolutionary Algorithms and MARL

Unmanned aerial vehicles (UAVs) are increasingly used to search and monitor operations for various tasks. However, their operational efficiency depends on advanced path planning that can balance conflicting objectives, such as minimizing mission time, maintaining connectivity to a ground control station (GCS), and prioritizing high-risk zones with time limitations. This paper formulates this challenge as a multi-objective optimization problem and provides a comparison between Evolutionary Algorithms (EAs) and Multi-Agent Reinforcement Learning (MARL). We implement and evaluate a Genetic Algorithm (GA), NSGA-II, NSGA-III, and an agent trained with Proximal Policy Optimization (PPO). Using a case study from a real-world risk map, we assess performance on mission duration, connectivity, and the speed of visiting priority and deadline-constrained regions. The results show that while EAs are computationally faster and effective for smaller grids, the PPO agent achieves superior performance, significantly reducing mission time while adjusting their trajectories for prioritized regions. Our analysis demonstrates the trade-offs between these methods and highlights the potential of MARL to develop more adaptive and effective disaster response systems.