## ompl-users

 [ompl-users] 2D Planner From: Kourosh Sartipi - 2013-10-26 12:31:02 Attachments: Message as HTML Hi everyone, This is my first time using ompl, so I apologize that my question is rudimentary. We have a robot that moves on the ground and want to do path planning using ompl and a 2D occupancy grid. Doing motion planning is very easy with RRTConnect, but our robot can only move forward and turn, so a chaotic motion is very costly. How can I tell ompl to favor straight lines? Since for every turn our robot has to completely stop, turn and start moving again. Thank you very much for your help Kourosh
 Re: [ompl-users] 2D Planner From: Mark Moll - 2013-10-26 19:38:26 Attachments: signature.asc Kourosh, You might want to look at the DubinStateSpace. This is an SE(2)-like state space for vehicles with a bounded turning radius that can only move forward. Interpolation between states is done using Dubins paths. If Dubins paths are not quite what you want, you can create your own state space. You could also play with the relative weight of orientation w.r.t. position in the distance function (to favor straight paths). Also, it’s recommended to use path simplification on a solution to mitigate some of the “chaotic motions” you have observed. Look at some of the demos for examples on how to do this. On Oct 26, 2013, at 7:30 AM, Kourosh Sartipi wrote: > This is my first time using ompl, so I apologize that my question is rudimentary. We have a robot that moves on the ground and want to do path planning using ompl and a 2D occupancy grid. Doing motion planning is very easy with RRTConnect, but our robot can only move forward and turn, so a chaotic motion is very costly. > How can I tell ompl to favor straight lines? Since for every turn our robot has to completely stop, turn and start moving again. -- Mark Moll
 Re: [ompl-users] 2D Planner From: Kourosh Sartipi - 2013-10-27 15:46:18 Attachments: Message as HTML Thanks Mark, I couldn't use DubinStateSpace or custom state spaces and my program would hang if I used them, but I guess it was the result of me being a beginner and messing something up somewhere. What did work however, was using PathSimplifier::reduceVertices. Now the planner gives me paths with as few vertices as possible and they are generally fine as long as I run path planner multiple times and select the best one. Thank you again, Yours, Kourosh On Sat, Oct 26, 2013 at 11:08 PM, Mark Moll wrote: > Kourosh, > > You might want to look at the DubinStateSpace. This is an SE(2)-like state > space for vehicles with a bounded turning radius that can only move > forward. Interpolation between states is done using Dubins paths. If Dubins > paths are not quite what you want, you can create your own state space. You > could also play with the relative weight of orientation w.r.t. position in > the distance function (to favor straight paths). > > Also, it’s recommended to use path simplification on a solution to > mitigate some of the “chaotic motions” you have observed. Look at some of > the demos for examples on how to do this. > > On Oct 26, 2013, at 7:30 AM, Kourosh Sartipi > wrote: > > This is my first time using ompl, so I apologize that my question is > rudimentary. We have a robot that moves on the ground and want to do path > planning using ompl and a 2D occupancy grid. Doing motion planning is very > easy with RRTConnect, but our robot can only move forward and turn, so a > chaotic motion is very costly. > > How can I tell ompl to favor straight lines? Since for every turn our > robot has to completely stop, turn and start moving again. > > -- > Mark Moll > > > >
 Re: [ompl-users] 2D Planner From: Ioan Sucan - 2013-10-29 08:38:04 Attachments: Message as HTML Hello Kourosh, Another thing you may find useful is the ParallelPlan and OptimizePlan classes, if you want to choose the best plan and have multiple plans computed in parallel http://ompl.kavrakilab.org/classompl_1_1tools_1_1ParallelPlan.html http://ompl.kavrakilab.org/classompl_1_1tools_1_1OptimizePlan.html Ioan On Sun, Oct 27, 2013 at 5:45 PM, Kourosh Sartipi wrote: > Thanks Mark, > I couldn't use DubinStateSpace or custom state spaces and my program > would hang if I used them, but I guess it was the result of me being a > beginner and messing something up somewhere. What did work however, was > using PathSimplifier::reduceVertices. Now the planner gives me paths with > as few vertices as possible and they are generally fine as long as I run > path planner multiple times and select the best one. > > Thank you again, > Yours, > Kourosh > > > On Sat, Oct 26, 2013 at 11:08 PM, Mark Moll wrote: > >> Kourosh, >> >> You might want to look at the DubinStateSpace. This is an SE(2)-like >> state space for vehicles with a bounded turning radius that can only move >> forward. Interpolation between states is done using Dubins paths. If Dubins >> paths are not quite what you want, you can create your own state space. You >> could also play with the relative weight of orientation w.r.t. position in >> the distance function (to favor straight paths). >> >> Also, it’s recommended to use path simplification on a solution to >> mitigate some of the “chaotic motions” you have observed. Look at some of >> the demos for examples on how to do this. >> >> On Oct 26, 2013, at 7:30 AM, Kourosh Sartipi >> wrote: >> > This is my first time using ompl, so I apologize that my question >> is rudimentary. We have a robot that moves on the ground and want to do >> path planning using ompl and a 2D occupancy grid. Doing motion planning is >> very easy with RRTConnect, but our robot can only move forward and turn, so >> a chaotic motion is very costly. >> > How can I tell ompl to favor straight lines? Since for every turn our >> robot has to completely stop, turn and start moving again. >> >> -- >> Mark Moll >> >> >> >> > > > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most > from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk > _______________________________________________ > ompl-users mailing list > ompl-users@... > https://lists.sourceforge.net/lists/listinfo/ompl-users > >
 Re: [ompl-users] 2D Planner From: Kourosh Sartipi - 2013-10-29 09:22:00 Attachments: Message as HTML Hi Ioan, Thanks for suggesting those classes, however it's better for me to not use multiple threads for planner, since I have other parallel threads running (visual odometry, mapping, etc.) and my laptop can't handle that kind of load. But for another project that needs offline planning I will certainly use those. Yours, Kourosh On Tue, Oct 29, 2013 at 12:07 PM, Ioan Sucan wrote: > Hello Kourosh, > > Another thing you may find useful is the ParallelPlan and OptimizePlan > classes, if you want to choose the best plan and have multiple plans > computed in parallel > > http://ompl.kavrakilab.org/classompl_1_1tools_1_1ParallelPlan.html > > http://ompl.kavrakilab.org/classompl_1_1tools_1_1OptimizePlan.html > > > Ioan > > > On Sun, Oct 27, 2013 at 5:45 PM, Kourosh Sartipi < > kourosh.sartipi@...> wrote: > >> Thanks Mark, >> I couldn't use DubinStateSpace or custom state spaces and my program >> would hang if I used them, but I guess it was the result of me being a >> beginner and messing something up somewhere. What did work however, was >> using PathSimplifier::reduceVertices. Now the planner gives me paths with >> as few vertices as possible and they are generally fine as long as I run >> path planner multiple times and select the best one. >> >> Thank you again, >> Yours, >> Kourosh >> >> >> On Sat, Oct 26, 2013 at 11:08 PM, Mark Moll wrote: >> >>> Kourosh, >>> >>> You might want to look at the DubinStateSpace. This is an SE(2)-like >>> state space for vehicles with a bounded turning radius that can only move >>> forward. Interpolation between states is done using Dubins paths. If Dubins >>> paths are not quite what you want, you can create your own state space. You >>> could also play with the relative weight of orientation w.r.t. position in >>> the distance function (to favor straight paths). >>> >>> Also, it’s recommended to use path simplification on a solution to >>> mitigate some of the “chaotic motions” you have observed. Look at some of >>> the demos for examples on how to do this. >>> >>> On Oct 26, 2013, at 7:30 AM, Kourosh Sartipi >>> wrote: >>> > This is my first time using ompl, so I apologize that my question >>> is rudimentary. We have a robot that moves on the ground and want to do >>> path planning using ompl and a 2D occupancy grid. Doing motion planning is >>> very easy with RRTConnect, but our robot can only move forward and turn, so >>> a chaotic motion is very costly. >>> > How can I tell ompl to favor straight lines? Since for every turn our >>> robot has to completely stop, turn and start moving again. >>> >>> -- >>> Mark Moll >>> >>> >>> >>> >> >> >> ------------------------------------------------------------------------------ >> October Webinars: Code for Performance >> Free Intel webinars can help you accelerate application performance. >> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most >> from >> the latest Intel processors and coprocessors. See abstracts and register > >> >> http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk >> _______________________________________________ >> ompl-users mailing list >> ompl-users@... >> https://lists.sourceforge.net/lists/listinfo/ompl-users >> >> >