Path planning of quality paths in the environment especially in the case of multi-objective path planning with obstacles poses a challenging task, where determining quality solutions can become a difficult challenge. Genetic algorithm and its heuristic search property offer a good solution for such problems. In the Master thesis, we present two ways of using genetic algorithms for the purpose of path planning.
The first method enables the search for piecewise linear paths in a static environment with obstacles considering the objective of path length and path smoothness. The algorithm utilizes a priori knowledge of the environment, which allows the use of dedicated genetic operators that contribute to the development of quality paths. To recognize obstacles in the environment we have developed a dedicated genetic operator capable of correcting paths that land within obstacles during the searching process of genetic algorithm. We presented the use of two different smoothness measures that allows multi-objective global path searching or the development of piecewise linear paths with a higher degree of smoothness.
The second method of using genetic algorithms describes the optimization of parametric curves that describe paths in space. Here we introduce two different methods for path smoothing and the genetic algorithm for both smoothing methods.
Through various tests, in the end, we confirmed that both methods develop quality paths. Nevertheless, due to the extent of possible solutions in more complex environments, genetic algorithm faces the requirements for a large population and high mutation probability, which leads to long computing times for the current implementation.
|
