I do believe your channel is brilliant and very informative . Videos you prepare are very consistent and perfectly done,. I am under big impression. I do believe you are going to achieve all your goals and success in each of your step. I keep finger for your path. Good luck!
Hello! It's completely correct, that the cost of each trajectory does not depend on the information from the costmap. However, the trajectories that go through or are quite close to the obstacles are directly rejected. The algorithm first generates the different trajectories. After generation, it rejects those trajectories that do not satisfy the velocity limits, curvature constraints and costmap constraints. Then from the remaining trajectories the best one is chosen based on the cost function. You can check the link for a smaller Python based implementation, which would be easier to understand: github.com/AtsushiSakai/PythonRobotics/blob/master/PathPlanning/FrenetOptimalTrajectory/frenet_optimal_trajectory.py