Menu
▾
▴
Re: [ompl-users] OMPL and real robot arm
|
From: Ioan S. <is...@gm...> - 2013-07-11 11:43:17
|
Hello Antoine,
Thank you! The best thing is to fork the ompl repository and send us a link
/ pull request to the changes. I think looking at the code will help me
better understand what the issue with distance / interpolate is.
Let us know if we can be of assistance!
Ioan
On Thu, Jul 11, 2013 at 10:57 AM, antoine aigueperse <a.a...@gm...
> wrote:
> Hello Ioan,
>
> Sorry, I forget to said the most important: I'm working only with revolute
> joints.
>
> To take into a count the bounds in functions distance() and interpolate(),
> i create a new state space based on the RealVectorStateSpace. If you want i
> can transfer you the class (or commit on your SVN)?
>
> Thanks,
>
> antoine
>
>
> 2013/7/8 Ioan Sucan <is...@gm...>
>
>> Hello Antoine,
>>
>> When you are calling functions from the state space, it is expected that
>> states are within bounds, except for calls to enforceBounds() and
>> satisfiesBounds().
>> So, calling the distance function with the value 231, when max is 230, is
>> considered invalid. For efficiency reasons, there is no check, but
>> behaviour is not defined.
>> I am not sure what you mean with 360 - 231 to be considered instead. Do
>> you mean that enforceBounds() should convert 231 to 360-231 (when max is
>> 230)? Or do you have revolute joints? If so, perhaps using SO2StateSpace
>> for some joints would help?
>>
>> To change the distance function you have to inherit from the
>> RealVectorStateSpace and provide your own implementation of distance(). It
>> sounds like that may be the case, if you would like distance() to behave in
>> a specific way. You would most likely need to implement interpolate() as
>> well.
>> I hope this helps. Please let me know if I did not understand your
>> question correctly.
>>
>> Ioan
>>
>>
>>
>> On Mon, Jul 8, 2013 at 7:26 PM, antoine aigueperse <
>> a.a...@gm...> wrote:
>>
>>> Hello OMPL users,I’m a beginner in OMPL, but this library looks very
>>> good and quite simple to use it (I hope for me!!).
>>>
>>> I have a question, maybe it is a stupid question and already solved in
>>> previous post. If it is the case sorry, I didn’t saw the post.
>>> I want to use OMPL to estimate path for an articulated robot arm having
>>> 7 DoF. I define the space which is composed to
>>> “RealVectorStateSpace(numjoints)” and joints bounds (see code below). Now
>>> I’m trying to compute distance between to ScopedState. My problem is that
>>> the distance cannot take into a count the joints bounds.
>>>
>>> For example the distance of the joint defined such as
>>> “armbounds.setLow(3, -130.0); armbounds.setHigh(3,230.0);” gives =>
>>> space->distance(129 , 231) = 102. Normally it should returned: abs(231-360)
>>> + abs(129) (distance respecting joint bounds). I think that, I will have
>>> the same problem for the interpolation.
>>>
>>> Is it possible to define a metric function (like
>>> si->setStateValidityChecker(boost::bind(&myStateValidityCheckerFunction,
>>> _1)); )? A solution is to create a complete state space using
>>> RealVectorStateSpace implementation?
>>>
>>> Thank in advance
>>>
>>> Antoine
>>>
>>> ________________________________
>>>
>>> #include <iostream>
>>>
>>> #include <ompl/base/spaces/RealVectorStateSpace.h>
>>> #include <ompl/geometric/planners/PlannerIncludes.h>
>>> #include <ompl/geometric/planners/rrt/RRTConnect.h>
>>>
>>>
>>> bool isStateValid(const ompl::base::State *state)
>>> {
>>> const ompl::base::RealVectorStateSpace::StateType *s =
>>> state->as<ompl::base::RealVectorStateSpace::StateType>();
>>>
>>> double X = s->values[0];
>>> double Y = s->values[1];
>>>
>>> if (X >=-170 && X <=-50 && Y >=0 && Y <=-30)
>>> return false;
>>>
>>> else
>>> return true;
>>>
>>> }
>>>
>>> void test(void)
>>> {
>>> /// construct the state space we are planning in
>>> int numjoints = 7;
>>> ompl::base::StateSpacePtr space(new
>>> ompl::base::RealVectorStateSpace(numjoints));
>>>
>>>
>>> ///We then set the bounds for the R^{numjoints} component of this
>>> state space:
>>> ompl::base::RealVectorBounds armbounds(numjoints);
>>>
>>> armbounds.setLow(0, -185.0); armbounds.setHigh(0, 0.0);
>>> armbounds.setLow(1, -155.0); armbounds.setHigh(1, 35.0);
>>> armbounds.setLow(2, 0.0); armbounds.setHigh(2,154.0);
>>> armbounds.setLow(3, -130.0); armbounds.setHigh(3,230.0);
>>> armbounds.setLow(4, -130.0); armbounds.setHigh(4,130.0);
>>> armbounds.setLow(5, -180.0); armbounds.setHigh(5,180.0);
>>> armbounds.setLow(6, -180.0); armbounds.setHigh(6,180.0);
>>>
>>> space->as<ompl::base::RealVectorStateSpace>()->setBounds(armbounds);
>>> space->setup(); // not necessary!!
>>>
>>> ///Create an instance of ompl::base::SpaceInformation for the state
>>> space
>>> ompl::base::SpaceInformationPtr si(new
>>> ompl::base::SpaceInformation(space));
>>>
>>> ///Set the state validity checker
>>> si->setStateValidityChecker(boost::bind(&isStateValid, _1));
>>> si->setup();
>>>
>>> ///Create a random start state:
>>> ompl::base::ScopedState<> start(space);
>>> start[0] = -0 ; start[1] = 0 ; start[2] = 0 ; start[3] = 0 ;
>>> start[4] = 129 ; start[5] = 0 ; start[6] = 0;
>>> std::cout<< "start.print() = " <<std::endl;
>>> start.print(std::cout);
>>>
>>> ///And a random goal state:
>>> ompl::base::ScopedState<> goal(space);
>>> goal[0] = 0 ; goal[1] = 0 ; goal[2] = 0 ; goal[3] = 0 ; goal[4] =
>>> 231 ; goal[5] = 0 ; goal[6] = 0;
>>> std::cout<< "goal.print() = " <<std::endl;
>>> goal.print(std::cout);
>>>
>>> // Check range axis
>>> std::cout<< "goal.satisfiesBounds = " <<
>>> space->satisfiesBounds(goal.get()) <<std::endl;
>>>
>>> // Compute L2 distance
>>> std::cout<< "goal.distance = " <<
>>> space->distance(start.get(),goal.get()) <<std::endl;
>>>
>>> std::cout<< "goal.enforcebounds = ";
>>> space->enforceBounds(goal.get());
>>> goal.print(std::cout);
>>> space->printSettings(std::cout);
>>>
>>> }
>>>
>>> int main()
>>> {
>>> test();
>>> }
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> This SF.net email is sponsored by Windows:
>>>
>>> Build for Windows Store.
>>>
>>> http://p.sf.net/sfu/windows-dev2dev
>>> _______________________________________________
>>> ompl-users mailing list
>>> omp...@li...
>>> https://lists.sourceforge.net/lists/listinfo/ompl-users
>>>
>>>
>>
>
|