vedmathai's profile - activity

If paged geometry were to be implemented in Gazebo then the possible objectives could be: Run an algorithm which determines the area of the world which has to be rendered and only render the models which directly exist in that portion. Thereby saving the amount of memory being used in one instance of running of the world. Is there some place where any detailed description of this exists? One good source is https://code.google.com/p/ogre-paged/ Also how would you page from the physics engine if you had to. (Nothing to do with the rendering engine). Is that even a good idea. Because lets say you have built a very big world and a model walking in another corner of it and is doing an action which we know will not have any effect on this model. Then that action should be ignored by the physics engine. Is it even a feasible idea?

yup it worked, thanks a lot.

and there is already an issue posted by John Hsu, https://bitbucket.org/osrf/gazebo/issue/607/link-getrelativeangularaccel-is-incorrect. But he was just referring to one of the functions. What I was asking was, why is it at every angular acceleration getter and setter function, only 'Torque/Mass' is used and not 'Torque/Inertia'.

The link is https://bitbucket.org/osrf/gazebo/src/585c79e153411203e6d7a26e98bfbc34afbc5319/gazebo/physics/Link.cc?at=default it is at line 578

Thanks for pointing out the mistake in the heading. I have edited that.

Correct me if I am wrong, but isn't AngularVelocity= Torque/ Moment of Inertia. Linear velocity = Force/ mass. Since both

math::Vector3 Link::GetRelativeLinearAccel() const { return this->GetRelativeForce() / this->inertial->GetMass(); }

and

math::Vector3 Link::GetRelativeAngularAccel() const { return this->GetRelativeTorque() / this->inertial->GetMass(); }

are using GetMass() then, both torque and force are being divided by mass to give you linear velocity and angular velocity respectively which I think is physically incorrect. But it has been done even in the setter functions under Link.cc, I don't know if it was wanted it that way. Or I am seeing it wrong.

I did not find any obvious answer anywhere for this, but is there any way we can just fill in equations instead of hard coded values into the sdf for the pose of different links on a model. As in there is a rod of length 5units and is tilted such that it makes 0.64 rads [arctan(3/4)] with the x axis. One end of the rod is at 0,0,0 and I want a globe to be created at the other end of the rod. Is it possible to do this by saying the globe should be positioned at (5cos(.64), 0, 5sin(.64)), instead of calculating and saying it should be (3,0,4) In this case it was easy as I took an example with number which could be handled at the back of an envelope but it does become little confusing when I have made this whole model with so many different links and suddenly after hard coding values in, I suddenly decide I need to change the height of one component. Then it becomes cumbersome to figure out all the values again. It seems to me that the entering of values in the sdf is quite brute force. Unless there is someother way to do it and I don't know. Thanks

Hi I am using ubuntu 12.04 and have installed gazebo by downloading the tarball file and following what was given on the install page(http://gazebosim.org/wiki/1.6/install) but when it comes to the last step: that is typing 'gazebo' into the terminal, this is what I get.

gazebo: error while loading shared libraries: libgazebo_common.so.1: cannot open shared object file: No such file or directory

I tried checking the answers on these pages http://answers.gazebosim.org/question... and http://answers.gazebosim.org/question...

Both of them had the same question, and their answers worked for the ones who had asked but not for me. As on the install page my installation was done at the default location, as I didn't specify anything.