The contribution of this paper is to extend existing sampling methods for the maneuver based trajectory generation for the use in large scale problems. The major problem which appear by the use of a purely or pseudo sampling method is that the computer spent its calculation time to calculate random numbers for possible points for tree expansion, which is generated by the maneuver based trajectory generation algorithm. The reason for this problem is that the time horizon which is used in the dynamic programming algorithm is to small to reach the random points out of the existing tree. To overcome this problem there exist some techniques like quasi random sampling. Here a dynamic method is proposed which allows to apply the pseudo sampling method for a large scale problem which at least drastically reduces the necessary calculation effort. It is also shown how the results can be applied for quasi random sampling.