Analysis and Prospect of Existing Path Planning Algorithms for Multi-Drone Systems

Drones are widely used nowadays in different sectors such as military, agriculture, surveillance, smart city measurements, logistics, and disaster rescue. In particular, multi-drone systems are gaining wider use for various problems, especially those requiring real-time navigation, simultaneous coordination, or collaboration between drones to attain various mission goals. One of the fundamental challenges for such systems is Path Planning, which requires computing an optimal collision-free trajectory from a source to a destination position in the presence of obstacles and moving agents. A wide range of algorithms, introducing nature-inspired metaheuristic methods like PSO, WOA, and LSO, have been developed to cope with this challenge. Reinforcement learning has also allowed drones to learn paths through environmental interaction and trial and error. Despite the strengths of those algorithms, they suffer from drawbacks, including computational complexity, scalability issues, being trapped in local minima, premature convergence, etc. This Review highlights the merits and demerits of existing Path Planning algorithms. It notes that while these algorithms have made considerable progress, they still suffer from limitations and do not yet account for the fractal and stochastic properties that appear in multi-drone systems. Fractal properties ensure solutions can adapt but remain valid and scalable throughout a hierarchy of environments. At the same time, stochastic methods enhance global exploration by weaving in a level of memory-less, random exploration of solutions. For future work, incorporating these features in the upcoming algorithms will make multi-drone performance even more robust, efficient, and adaptable and encourage their successful use in the real world.