A UAV path planning with parallel ACO algorithm on CUDA platform

Solving the path planning problem of a UAV is a challenging issue especially if there are too many checkpoints to visit. Mainly, the brute force approach is used to find the shortest path in the mission area, which requires too many times to find a solution. Therefore, evolutionary algorithms and swarm intelligence techniques are used to find a feasible solution in an acceptable time. In this study, path planning problem of a UAV is solved by using a highly parallelized Ant Colony Optimization (ACO) algorithm on CUDA platform. The UAV path is constructed for disseminating keys and collecting data from a Wireless Sensor Network, which is previously defined. Due to its simplicity and effectiveness, ACO is selected as a path planning algorithm. However, ACO is not satisfactory if the mission area becomes large and there are an excessive number of checkpoints and/or additional constraints. In order to increase the performance, some parallelization techniques must be used in high performance computing platforms. GPU architecture has emerged as a powerful and low cost architecture for enabling impressive speedups for scientific calculations. Therefore, the parallel structure is constructed on CUDA architecture. The experimental results are compared with the CPU performance of the serial algorithm, and they clearly show that the proposed approach have a great potential for acceleration of ACO and allow to solve many complex tasks such as UAV path planning problem. We also present the execution results with different parameter values to expose the results for the researchers.