Traditional path planning and obstacle avoidance of discontinuities often exist. This paper presents a new obstacle avoidance path planning for a smooth approach-genetic algorithms and B-spline curve. First of all, first through the collision detection, can detect obstacles in the path ahead, then the genetic algorithm combined with B-spline curve and then map out a smooth obstacle avoidance path. The algorithm can avoid skidding caused by movement and target location error, solve the robot in a static environment, global, local path planning and path planning problem of discontinuity exists. Simulation and experimental results show the effectiveness of the algorithm.