Collaborative Path Planning and Collision Avoidance for Multi-UAV Navigation based on Accelerated Improved Particles Swarm Optimization

In recent years, multi-UAV systems (MUS) have raised more and more attention due to their advantages in time efficiency, complementarity, and flexibility compared to a single UAV. Multi-UAV path planning is a fundamental problem that is necessary for MUS navigation and finding non-collision trajectories at the same time. In this work, we study the current challenge of multiple multirotor UAV path planning and propose an innovative algorithm and vehicle-to-vehicle (V2V) decentralized communication architecture that is suitable for real-time multirotor UAV navigation without Global Navigation Satellite Systems (GNSS). Existing algorithms and research are mostly finished on simulations and lack real flights to verify feasibility. Specifically, in order to achieve real-time applications, the proposed algorithm is based on Particle Swarm Optimization (PSO) due to its scalability and easy-to-implement, dubbed Accelerated Improved Particle Swarm Optimization (AIPSO). Nevertheless, standard PSO has two obvious drawbacks that might compromise our goal, slow convergence rate and easy falling into local optimal trajectories. The novelty in AIPSO overcomes two aforementioned PSO problems by introducing the Simulated Annealing (SA) algorithm and Dimensional Learning Strategy (DLS). Moreover, we restrain the number of particles to accelerate computations. Lastly, we integrated AIPSO with a decentralized communication architecture, called Decentralized Multi-UAV AIPSO (DMU-AIPSO). In order to verify the feasibility of DMU-AIPSO, both simulations and real flights are presented.