The optimal path search is important in navigation learning with an electrical wheelchair. This paper studies six basic approaches for path planning in a static environment: breath-first search, depth-first search, A*, Moore-Dijkstra, neural approach and genetic algorithms. Firstly, the environment is modeled in a 2D grid and the genetic approach has fixed chromosome length. Then the genetic approach is evaluated within a grid free environment and with variable chromosome lengths. The latter approach offers good solution in precision and in computation time, under certain conditions. Tests are led and the results are discussed through experiments.