Abstract: Combining the curvature continuous feature of cubic B-spline and the superior global search performance of genetic algorithm, a docking path planning method suitable for underactuated AUV (autonomous underwater vehicle) is designed. A smooth 3-dimensional path from the start point to the end point is given for the homing stage in AUV recovery, by which appropriate AUV position and attitude required by the following guided docking stage are guaranteed. Firstly, geometrical constraints caused by underactuated AUV are analyzed, including terminal constraint and kinematical constraint. Secondly, an idea of determining 3-D path by selecting control points sequence is presented according to the features of B-spline curves. In the first step, some control points are selected through analyzing mission terminal constraints in template pattern. In the second step, the control points in the middle section are given out after adaptive heuristic search in the solution space by using gene algorithm, whose fitness function contains constraints of AUV motion like rotation and heaving respectively in horizontal and vertical plane. The curve defined by the previous two sections of control points finally satisfies all geometrical constraints. At last, semi-physical dynamical simulation experiments on path generation and tracking are carried out. The tracking result shows that the geometrical character of path matches the maneuverability of AUV. Position and attitude of AUV are suitable for the following guided docking stage while path tracking finished.