<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body dir="auto"><div>Andreas,</div><div id="AppleMailSignature"><br></div><div id="AppleMailSignature">My experience with openrave kinematics has been with ikfast. Ikfast first runs offline on a model to create a "closed form" IK solution. This really only works with 6-DOF and lower chains. Additionally, I've found that ikfast solutions are generally not very precise, including providing answers near but still slightly outside joint limits and providing joint configuration answers that when pushed back through the forward kinematics are not very close to the original desired Cartesian pose. Thus, in the past when I've used ikfast, I also end up having to run KDL (or trac-IK) afterwards to quickly coverage to a solution with small (< 1e-2) residual error wrt the desired Cartesian pose. Finally ikfast does not allow tolerances for Cartesian dimensions, which allows things like position only solutions or for example solutions where end effector roll can fall within a range.</div><div id="AppleMailSignature"><br></div><div id="AppleMailSignature"><br>Sent from my iPhone</div><div><br>On Nov 5, 2015, at 3:47 PM, andreas.tenpas--- via ros-users <<a href="mailto:ros-users@lists.ros.org">ros-users@lists.ros.org</a>> wrote:<br><br></div><blockquote type="cite"><div>This is neat! Thanks for sharing!<br><br>I would be interested in how this performs in comparison to openrave.<br><br>Best,<br><br>Andreas<br><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"><div class="quote" style="line-height: 1.5"><br><br>-------- Original Message --------<br>Subject: [ros-users] Introducing A Better Inverse Kinematics Package<br>From: Patrick Beeson via ros-users <ros-users@lists.ros.org><br>To: <a href="mailto:ros-users@lists.ros.org">ros-users@lists.ros.org</a><br>CC: <br><br><br type="attribution"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">
<div bgcolor="#FFFFFF" text="#000000">
<div dir="ltr">
<div>TRACLabs Inc. is glad to announce the public release of our
Inverse Kinematics solver TRAC-IK. TRAC-IK is a faster,
significantly more reliable drop-in replacement for KDL's
pseudoinverse Jacobian solver.<br>
<br>
Source (including a MoveIt! plugin) can be found at:<br>
<a href="https://bitbucket.org/traclabs/trac_ik.git" target="_blank">https://bitbucket.org/traclabs/trac_ik.git</a><br>
<br>
</div>
<div>TRAC-IK has a very similar API to KDL's IK solver calls,
except that the user passes a maximum time instead of a maximum
number of search iterations. Additionally, TRAC-IK allows for
error tolerances to be set independently for each Cartesian
dimension (x,y,z,roll,pitch.yaw).<br>
</div>
<div><br>
</div>
More details:<br>
<div>
<div><br>
KDL's joint-limited pseudoinverse Jacobian implementation is
the solver used by various ROS packages and MoveIt! for
generic manipulation chains. In our research with Atlas
humanoids in the DARPA Robotics Challenge and with NASA's
Robotnaut-2 and Valkyrie humanoids, TRACLabs researchers
experienced a high amount of solve errors when using KDL's
inverse kinematics functions on robotic arms. We tracked the
issues down to the fact that theoretically-sound Newton
methods fail in the face of joint limits. As such, we have
created TRAC-IK that concurrently runs two different IK
methods: 1) an enhancment of KDL's solver (which detects and
mitigates local minima that can occur when joint limits are
encountered during gradient descent) and 2) a Sequential
Quadratic Programming IK formulation that uses quasi-Newton
methods that are known to better handle non-smooth search
spaces. The results have been very positive. By combing the
two approaches together, TRAC-IK outperforms both standalone
IK methods with no additional overhead in runtime for small
chains, and significant improvements in time for large chains.<br>
<br>
</div>
<div>Details can be found here in our Humanoids 2015 paper here:<br>
<a href="https://personal.traclabs.com/%7Epbeeson/publications/b2hd-Beeson-humanoids-15.html" target="_blank"></a><a href="https://personal.traclabs.com/~pbeeson/publications/b2hd-Beeson-humanoids-15.html" target="_blank">https://personal.traclabs.com/~pbeeson/publications/b2hd-Beeson-humanoids-15.html</a><br>
<br>
</div>
<div>A few high-level results are shown in the attached
(low-res) figure.<br>
</div>
<div><br>
</div>
</div>
</div>
</div>
</div>
</blockquote></ros-users@lists.ros.org></div></div></blockquote><blockquote type="cite"><div><span>_______________________________________________</span><br><span>ros-users mailing list</span><br><span><a href="mailto:ros-users@lists.ros.org">ros-users@lists.ros.org</a></span><br><span><a href="http://lists.ros.org/mailman/listinfo/ros-users">http://lists.ros.org/mailman/listinfo/ros-users</a></span><br></div></blockquote></body></html>