ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

ompl - planning under differential constraints

asked 2013-05-14 11:10:45 -0600

pacifica gravatar image

Hi guys!

Anyone had never used the ompl ( library to plan the motion of a mobile robots ( car-like or quadrotor )?? I've read all tutorials and demos too, but, while with the geometric constraint the result are really interesting, under differential constraints the results are unusable, the planned trajectory are always approximate or too much inaccurate. I've the same results with the appgui... anyone can tell me if i wronging something? Anyone have good results??

Thanks a lot.

(NB i'm using the last version of ompl in standalone version).

edit retag flag offensive close merge delete


Tell me if you are still interested in a response to this question.

Javier V. Gómez gravatar image Javier V. Gómez  ( 2015-06-11 10:29:36 -0600 )edit

I'm interested :)

kwiecien gravatar image kwiecien  ( 2015-11-09 08:26:23 -0600 )edit

1 Answer

Sort by » oldest newest most voted

answered 2015-11-09 10:45:00 -0600

OMPL planners under differential constraints are not optimal. I, together with OMPL developers, started an effort to support optimal planners with differential constraints but this is requires a huge effort (to do it in a hacky way is not that much, I did it in the past but I think I deleted the branch :( ) and also gives some problems that it is, IMO, not really clear how to deal with them (such as system simulation and 2 point boundary value solvers accuracy). Actually, this is an state of the art problem, not OMPL problem.

The consequence of using non-optimizing planners is that the plans are usually bad. For instance, for a car-like robot it will do lots of maneuvers.

However, the question refer to approximate results. If you mean OMPL approximate solution (there is a path but it does not reach the goal region), then most likely you are doing something wrong: bad parameters, wrong problem formulation, etc. For me it has worked in the past.

Also it is mention that the trajectories are "too much inaccurate". With respect to what? Most likely the OP refers to the optimality of the path (reason already explained).

edit flag offensive delete link more

Question Tools


Asked: 2013-05-14 11:10:45 -0600

Seen: 859 times

Last updated: Nov 09 '15