Meteorological and Topographical Big Data-Driven UAV Trajectory Planning

We propose a comprehensive trajectory planning framework for quadcopter unmanned aerial vehicles (UAVs) assisting in edge computing, aiming to control their crash proba-bility when deployed at the edge of the Internet of Things (IoT) across diverse meteorological and topographical environments. First, we collect hourly meteorological data from the National Oceanic and Atmospheric Administration's (NOAA) Integrated Surface Data (ISD), which includes 16 key metrics such as temperature, atmospheric pressure, wind speed, and solar radiation. We then use the Spark framework and a Long Short-Term Memory (LSTM) network, accounting for seasonal trends, to predict meteorological conditions based on the collected data. Subsequently, we combine a dynamic risk assessment model with the Endoh crash risk model to calculate the probability of a crash. Finally, we design a three-dimensional ant colony system variant, referred to as ACS-DS-R, to plan UAV trajectories, incorporating meteorological predictions and topographical information. Our proposed framework dynamically adjusts UAV paths based on terrain data and real-time meteorological predictions to avoid high-risk areas, significantly reducing crash rates. We present extensive numerical results using real data from across Japan-including cities, forests, lakes, coasts, and mountainous areas-demonstrating that our approach reduces crash probabil-ity by more than 33%, improves the task completion rate by about 2.58 times while increasing total energy consumption by only 13% compared to existing mainstream approaches. Our work demonstrates the value of integrating real-time meteorological predictions and topographical data into UAV trajectory planning, offering a practical solution for safer and more efficient UAV deployments in diverse environments.