RelaxedIK Solver
DEVELOPMENT BRANCH
NOTE: The dev branch is not quite ready for public use. I'm still testing out many of the new features, and the documentation has not been updated to reflect the changes. I would recommend sticking with the master branch for now - I'll be merging the dev branch to master when it's all ready to go in a few weeks
Welcome to RelaxedIK! This solver implements the methods discussed in our paper RelaxedIK: Real-time Synthesis of Accurate and Feasible Robot Arm Motion (http://www.roboticsproceedings.org/rss14/p43.html)
Video of presentation at RSS 2018 (RelaxedIK part starts around 12:00) : https://youtu.be/bih5e9MHc88?t=737
Video explaining relaxedIK https://youtu.be/AhsQFJzB8WQ
RelaxedIK is an inverse kinematics (IK) solver designed for robot platforms such that the conversion between Cartesian end-effector pose goals (such as "move the robot's right arm end-effector to position X, while maintaining an end-effector orientation Y") to Joint-Space (i.e., the robot's rotation values for each joint degree-of-freedom at a particular time-point) is done both ACCURATELY and FEASIBLY. By this, we mean that RelaxedIK attempts to find the closest possible solution to the desired end-effector pose goals without exhibiting negative effects such as self-collisions, environment collisions, kinematic-singularities, or joint-space discontinuities.
To start using the solver, please follow the step-by-step instructions in the file start_here.py (in the root directory)
If anything with the solver is not working as expected, or if you have any feedback, feel free to let us know! (email: rakita@cs.wisc.edu, website: http://pages.cs.wisc.edu/~rakita) We are actively supporting and extending this code, so we are interested to hear about how the solver is being used and any positive or negative experiences in using it.
Citation
If you use our solver, please cite our RSS paper RelaxedIK: Real-time Synthesis of Accurate and Feasible Robot Arm Motion http://www.roboticsproceedings.org/rss14/p43.html
@INPROCEEDINGS{Rakita-RSS-18,
AUTHOR = {Daniel Rakita AND Bilge Mutlu AND Michael Gleicher},
TITLE = {{RelaxedIK: Real-time Synthesis of Accurate and Feasible Robot Arm Motion}},
BOOKTITLE = {Proceedings of Robotics: Science and Systems},
YEAR = {2018},
ADDRESS = {Pittsburgh, Pennsylvania},
MONTH = {June},
DOI = {10.15607/RSS.2018.XIV.043}
}
If you use our solver for a robot teleoperation interface, also consider citing our prior work that shows the effectiveness of RelaxedIK in this setting:
A Motion Retargeting Method for Effective Mimicry-based Teleoperation of Robot Arms https://dl.acm.org/citation.cfm?id=3020254
@inproceedings{rakita2017motion,
title={A motion retargeting method for effective mimicry-based teleoperation of robot arms},
author={Rakita, Daniel and Mutlu, Bilge and Gleicher, Michael},
booktitle={Proceedings of the 2017 ACM/IEEE International Conference on Human-Robot Interaction},
pages={361--370},
year={2017},
organization={ACM}
}
An Autonomous Dynamic Camera Method for Effective Remote Teleoperation https://dl.acm.org/citation.cfm?id=3171221.3171279
@inproceedings{rakita2018autonomous,
title={An autonomous dynamic camera method for effective remote teleoperation},
author={Rakita, Daniel and Mutlu, Bilge and Gleicher, Michael},
booktitle={Proceedings of the 2018 ACM/IEEE International Conference on Human-Robot Interaction},
pages={325--333},
year={2018},
organization={ACM}
}
Dependencies
kdl urdf parser:
sudo apt-get install ros-[your ros distro]-urdfdom-py
sudo apt-get install ros-[your ros distro]-kdl-parser-py
sudo apt-get install ros-[your ros distro]-kdl-conversions
readchar:
sudo pip install readchar
fcl collision library: https://github.com/BerkeleyAutomation/python-fcl
sudo pip install python-fcl
scikit learn: http://scikit-learn.org/stable/index.html
sudo pip install scikit-learn
scipy:
sudo pip install scipy
Pyyaml:
sudo pip install PyYaml
To use the Julia version of the solver (which is the recommended option), you will first need to install Julia. https://julialang.org/
The solver was written in Julia 1.0.2, though any more recent 1.X.X version should suffice.
If you plan to extend any of the Julia code, we recommend using the Juno IDE (not required)
Once Julia is installed, initialize a Julia environment using the following command:
julia
then run the following commands to install Julia dependencies for RelaxedIK:
using Pkg
Pkg.add(["YAML", "BenchmarkTools", "ForwardDiff", "Calculus", "ReverseDiff", "StaticArrays", "Rotations", "Flux", "BSON", "NLopt", "Knet", "Random", "RobotOS", "Distributions", "PyCall", "Dates", "LinearAlgebra"])
in the same Julia session, configure PyCall within Julia by running the following commands:
ENV["PYTHON"] = "/usr/bin/python2.7"
Pkg.build("PyCall")
Tutorial
For full setup and usage details, please refer to start_here.py in the src directory. Prior to starting the setup process, please ensure that you have installed all dependencies listed above.
Usage Notes
If you are using the Julia version of the RelaxedIK solver, note that it currently takes a little while (~30 seconds) for the solver to start. Julia code is VERY fast once it starts up, but it takes a bit of overhead time to do its JIT compilation.
Unfortunately, there's little we can do at this point to eliminate this JIT compilation time the first time RelaxedIK starts up (though, we are hoping compilation times get better in general in future versions of the Julia programming language!). However, to alleviate this issue in the short term, we have provided a way to "reset" the solver on subsequent runs, such that you will not have to kill the program, restart it, and wait another 30 seconds for it to compile every time.
To perform this reset, first start the following ROS node using this command:
rosrun relaxed_ik reset_and_quit.py
Then, make sure that the terminal that that command was entered into has focus (i.e., it has been clicked), and type the character 'r'. This will reset the robot to its initial configuration. If you would like to quit instead of resetting, type the character 'q'.
Coming Soon
RelaxedIK has been rewritten in Julia to substantially boost performance. This version of the solver is currently in a beta test on this branch. If you have feedback based on your experience using this version of the solver (positive or negative), please email me at rakita@cs.wisc.edu. Thanks!
Change Log
Version 1.1 (8/16/18): added support for prismatic and fixed joints