In the trajectory planning task of autonomous driving, we usually use center line of the road as a reference line to help planning. However, the curvature may be large in curve lanes. So I build this project learning from karlkurzer/path_planner. There are Hybrid A* searching, smoothing, Rviz visualization in his project but I'm mainly focus on smoothing. The main changes are the calculation of smooth cost and voronoi cost. And I use OpenCV to visualize the result.

I put an example to generate image maps, "generate_lane_curves". And another more important example to test the method, "path_smoother_example".

If you want a good result, some adjustment for parameters is needed.

A typical result is show below. Green curve is the original path without smoothing, actually it is the center line plus a random noise. Red curve is the smoothed path.