Hybrid Offline-Online UAV Optimal Path Planning and Outbreak Dynamic Autonomous Behavior

This paper presents a novel hybrid path planning framework designed for autonomous Unmanned Aerial Vehicles (UAVs) operating in dynamic and uncertain environments. The proposed approach integrates an Offline Phase that leverages a Genetic Algorithm (GA) to optimize PID control parameters and velocity profiles, alongside an A* search algorithm for initial path generation on static obstacle maps. This phase establishes an energy-efficient and optimized baseline trajectory. The Online Phase is activated only upon the detection of unexpected events or dynamic obstacles. Here, a Parallel Probabilistic Cellular Automata with Monte Carlo Sampling (P-PCA-MCS) system is employed for real-time collision avoidance. This system dy- namically updates and fuses PCA-based occupancy probabilities with Monte Carlo-sampled collision probabilities for adversarial drone trajectory prediction, resulting in a comprehensive risk map. At predefined replanning intervals, the drone evaluates motion primitives based on a quality function derived from these fused probabilities, enabling rapid and adaptive trajectory adjustments to avoid dynamic threats while striving to return to the pre-optimized path. Extensive simulations across vary- ing complexities demonstrate that the P-PCA-MCS algorithm consistently achieves superior performance. Compared to other state-of-the-art methods, it significantly reduces collision rates, maintains near-optimal path efficiency, and exhibits remarkably low computation burden, proving its efficacy for robust, real-time autonomous drone navigation in high-density airspaces.