Abstract: The traditional genetic algorithm has the faults of slow evolving speed being easy
to get into local optimum etc., an improved one based on it is provided. The improved one extends
the barriers based on the principle of Minkowshi theory in the step of estimating the path and
improves the graphic theory to construct the real feasible region. In the step of calculating the
shortest path, population initialization individual coding method and so on are researched and the
concept of similarity in the selection operator is imported which can enlarge the range of the early
population and to avoid to get into the local optimum. At last, simulation test proves that the
improved algorithm is feasible.

Key words: computer application, genetic algorithm, mobile robot, path planning