Metaheuristic Optimization for Path Planning in UAV Networks for Long-Distance Inspection Tasks

With broad applicability in surveillance, delivery, and environmental monitoring, multi-drone path planning has received a great deal of attention recently. So far, wide applicability has been explored by various applications of Genetic Algorithms, Particle Swarm Optimization, and Ant Colony Optimization, which have already established reliable solutions to navigation and coverage through drone path optimization. However, in dynamic, obstacle-dense environments, these algorithms fail to adapt in real-time with energy efficiency. Traditional GA and PSO approaches can be effective for generating an initial path but fail to respond to collision avoidance and efficient resource management. In complex and unpredictable conditions, ACO fails to converge rapidly, which leads to suboptimal paths and increased energy consumption. To overcome such barriers, we present here an HGARL for enhanced multi-drone path optimization. The proposed algorithm thus combines the population-based search capability of GA with RL's adaptive learning ability, therefore providing an algorithm that in addition to finding optimal paths adjusts the paths according to changing environmental situations. HGARL presents better performance compared to existing methods. In path convergence, it reaches up to 15%, and in energy consumption, it reduces up to 20% with safe collision-free navigation. The strong potential of HGARL for robust, efficient, and adaptive solutions of real-time multi-drone operation in complex and dynamic environments has set a new benchmark for UAV path planning.