Best tool for motion planning of autonomous vehicles?
I am planning to simulate some common motion planning algorithms on a car - including sensor readings in a virtual environment, applying the algorithm, smoothing the best trajectory and performing the low-level control.
I have decided that using OMPL will be the most prudent course, but I am unsure which tool to use with it. I've read many sources which have used MoveIt, MORSE, etc. interfaced with OMPL to simulate the motion. Please can you suggest which would be the best option for a car?
I've read about teb_local_planner as well but I'm not exactly sure, is it meant to be used in conjunction with a global motion planning algorithm like RRT? Because the description says that it optimizes the trajectory generated by a global planner.
Any help in this regard will be much appreciated.
Have you looked into using the navigation stack with move_base? It is intended for mobile robots whereas MoveIt! is intended for end effector planning.
I need to implement a randomized motion planning algorithm like RRT. move_base only supports algorithms like Djikstra and A*. That's why I was planning on using the OMPL. But I'm still not sure what to use OMPL with on ROS.
teb_local_planner does not require a global plan that is aware of robots kinematic constraints. The default global planner works well. However, I have not tried the application to cars, the package is tuned for fairly slow robots and not cars with high speeds (->exploiting dynamics). But it may work
Thanks for the comment! I've spent quite some time trying to understand teb_local_planner and the global_planner. I need to implement RRT* specifically for my robot, so I looked at the option of writing a global planner plugin. The srl_global_planner package exists on github but it gives errors(1/2)
So I've decided to write my own global planner plugin. For this I wanted to ask, is it possible to use ompl to write it? Also, is the obstacle avoidance considered at the global planner or local planner level? (2/2)