Robotics StackExchange | Archived questions

What planning algorithm is the pose_follower package using?

I am just curious what kind of local planner the pose_follower uses. I saw a post saying that it's a very basic algorithm. What exactly is the algorithm? Is there a paper?

Asked by RicoJ on 2020-07-17 14:57:32 UTC

Comments

Answers

Ok, I looked through the code, Here is what I find, let me know if there's anything off!

Basics: calculates the twist between the current pose and the next waypoint. Then, make simple modifications to this twist and use it as cmd_vel

1. This planner will prefer turn and go if the user explicitly asks so and the angle in the twist is smaller than a threshold. 

2. If the twist's linear/angular velocity is too high or too low, it will simply scale it down to the speed limit.  This capping might be problematic as it changes the arc of the planner, e.g, If the angular component is large, linear component is small, the planner might end up outputting a series of crazy arcs.  

3. Then if the current pose and the target pose is too close, the  local planner will check the next goals to see if goal has been reached 

4. If a new global plan is not ready yet, the local planner will keep planning for the existing plan.  

Asked by RicoJ on 2020-07-17 16:39:44 UTC

Comments

That's it, there's no real planning going on, I would call it a controller more than a planner ;)

Asked by achille on 2020-07-19 16:33:40 UTC