Path Planners Comparison

Hi @staff I would like to ask gently for an advice in the following topic:

  • I wish to write an article comparing path planners efficiency and features when different planners (Global and/or Local) are applied to the same scenario. I have already read some papers and material on internet, and the researchers compare them usually in the following aspects:

1 - Computational Cost
2- Sharp Obejects Detection
3- Behavior when facing Dynamic obstacles
4 - Trajectory Smoothness
5 - Path Length
6 - Time of fly the robot takes to execute the generated path
7 - Clearance (I confess I did not understand what this means, please tell me if you know)

Well these were the features I have read in a fast survey…
Now comes the ROS Question, hehehe.

**A) Are there some ready available tools in ROS to measure these features for later comparison? considering I am using the Navigation Stack (not MoveIt or other framework, just using the move_base node, gmapping, amcl…the standard way) for a self-driving car simulation?
If there are not available tools:
B) how would you and how can I make a kind of measure of these parameters, I have briefly described previously (1-7), in ROS?

C) Would you suggest me other parameters used to compare planners (besides the ones I had written 1-7)? and how can I compare them in ROS ?

D)In a summarize what makes sense to be taken into account when compating path planners in order to choose the most suitable one ?( I mean for a simple case of evasive maneuver only facing a static or dynamic object, not complex scenarios)

E)And should I and Can I compare the Global and Local planner using different comparison metrics or not?

F) Additional Questions: **
** F.1)Is there some repository with all available path planners for ROS melodic distro? I would like to use the RRT path planner for my simulation project, however I have read it was developed to ROS kinetic **
** F.2) Is it possible to “quick” adjust them for other ROS version or does this take too time?"

It is not a critic, because I really liked very much the Path Planning Modules and the video class from @albertoezquerro. They are really excelent and self-explanatory. It is just a suggestion for maybe in future you could add an extra class in the end of this module…of metrics to choose the most suitable planner for an standard application (drone, robot, car)…because a ROS student beginner (my case) can use a planner but don t know the reasons why he/she has chosen this one.

Thanks in advance

Best Regards

Marcus

1 Like

Hello @marcusvini178,

Well, this sounds really interesting. In fact, I would suggest you create a ROSject (or even a course) for running these experiments. This way, whenever you share your article, anybody will be able to reproduce your experiments with just a simple click.

Unfortunately, I’m afraid this goes past my current knowledge on ROS Navigation. Because of this, I would suggest you post this question in the ROS Answers forum, where you will probably find people who will be able to give you better advice than me on this issue.

Best,

1 Like