<![CDATA[Autonomous mini unmanned aerial vehicles (UAVs) for civilian applications face a critical challenge: during flight, their mission planning cannot break down complex goals into real-time actions. It’s like having a brilliant strategy with no way to execute it in the moment conditions change. While current solutions can handle basic navigation, they often fail when conditions change. This lack of adaptability seriously limits autonomy in real-world applications, like infras tructure inspection or emergency response. The core problem? Nobody has yet built a system that can think in both layers, combining hierarchical goal decom positions with dynamic tasks without overloading the onboard computer. Our work addresses this gap by introducing an integrated mission planning system with two complementary modules. First: the decision-making module employs recursive goal tree construction to transform high-level mission goals into hier archical sub-goal structures in a systematic manner. Second: the task planning module converts these structured goals into concrete MAVLink command se quences. Together, these modules bridge the gap between abstract mission spec ifications and low-level flight operations while enabling dynamic replanning. To verify if our system actually works, we validated the framework through simulation-based experiments using a Python UAV mission simulator across 50 test runs. The results showed a 94% mission completion rate, with an average planning time of 1.8 seconds for missions with 5 to 8 waypoints. It adapted well to surprises: new targets (100% success), no-fly zones (92% success), and priority changes (96% success). Compared to traditional reactive baseline ap proaches, the framework reduced replanning time by 67%. This tells us that the modular approach is not just theoretically sound but it’s also practically viable for real-world civilian operations.]]>
