| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 101 === In this study, an improved non-dominated sorting genetic algorithm (improved NSGA-II) is proposed for multi-UAVs three-dimensional path planning. It can be found that the problem for such a path planning problem is a multiple objectives optimization problem in a 3-D real environment. In addition to the use of real maps as the flight environment, several restrictions and the possibility of destruction and detection caused by protective mechanisms (ADUs) are also considered in the problem. In order to have better search performance, in addition to using the original environment of natural selection methods, humankind selection manner is taken into account to determine the direction of genetic evolution by human preference. This approach simultaneously considers the weight assignments of multiple objectives and reduces the costs of convergence time, which may increase if the fitness functions chosen are improper. Also, with a fixed horizontal coordinate in the three-dimensional coordinates, the algorithm does not converge to the same point so that it can reach feasibility of the paths. Besides, new designed genetic operators, and , are proposed to increase the rationality and diversity of paths. Experimental results indicate that the proposed algorithm can take several objectives into account and effectively find feasible solutions of the approximate Pareto optimum set based on the human preferences.
|
|---|
