kevint1989/robotics-toolbox-python

Robotics Toolbox for Python - Update 1

robotics-toolbox-python-update1

Robotics Toolbox for Python - Update 1

This is a 1st update on the 'old first attempt' at creating an open Python version of Peter Corke's Robotics Toolbox for MATLAB^® (RTB) © 1993-2018. Although the Robotics Toolbox for Python © 1993-2016 repository has languished over the past decade, there may still be some interest in its revival; even in its dated and rudimentary form. At this point, there is no intent to morph RTB for Python - Update 1 into another feature rich robot simulation tool extensively dependent on 3rd party packages and libraries. However, it should be noted that the present RTB for Python architecture relies exclusively on the 'pylab' convenience module, a practice not recommended by Matplotlib developers.

For those interested in an active Robotics Toolbox for Python port, Peter Corke directs your attention to robopy.

Getting Started

Download, unpack and run demos and tests as described in the README files, or optionally just build and install with the provided setup.py script (e.g., python setup.py install). Best use experience with RTB for Python is in IPython's interactive Pylab shell mode. For example, to display an animation of the Puma 560 (standard) robot transition from a nominal pose to a ready pose, invoke IPython in the demo or test directory and enter the six commands as shown below.

$ ipython --matplotlib

In [1]: %pylab

In [2]: import _robot

In [3]: from robot import *

in [4]: from robot.puma560 import *

In [5]: (Q,_,_) = jtraj(qn,qr,51)

In [6]: rbplot(p560, Q)

The above will generate an animation of a stick figure rendering of the Puma 560 as depicted in the image below. Clicking on the image will bring up an MP4 video clip of the animation created with 'rec=1' passed as a third argument to rbplot() on input line [6] above.

Several Jupyter notebook examples of RTB for Python are presented in the Robotics Toolbox for Python - Update 1 repository Wiki on GitHub. These notebooks may be viewed online by selecting the associated [nbviewer] link. Consult the README.txt file in the notebooks directory for instructions on running Jupyter IPython notebook examples and corresponding Python source files locally.

Guidelines for Update 1

1. Minimize changes to facilitate code comparison against sourced Python and MATLAB^® scripts.
2. Maintain the 'look and feel' of the original RTB functional interface.
3. Preserve content of function's descriptive comment blocks, but modify to reflect code changes and adhere to original epydoc docstring conventions.
4. Retain copyright, license and disclaimer notices.

Accomplishments for Update 1

1. RTB for Python is now compatible with Python 2 and 3 variants.
2. Revitalized forward and inverse dynamics functionality.
3. Incorporated capability to display and record simple animated 2D and 3D plots of frame transforms and robot manipulator poses.
4. Extended functionality to run additional demo and test cases provided in RTB for Python and MATLAB^® distributions.
5. Updated inverse kinematics in RTB for Python to match RTB version 9.8.
6. Fixed several code errors, and made some minor improvements -- notably in parsedemo and testparser.
7. Corrected semantic and syntactic mistakes in some comments.
8. Removed insidious leading tabs and unnecessary terminating semi-colons from executable statements when encountered during code editing.

Although this effort was primarily conducted on an Ubuntu 14.04 x86_64 Linux system with the Python, NumPy, Matplotlib and Octave versions identified below, the updated RTB for Python should work on non-Linux systems which support Python 2 or 3, and SciPy.

• Python 2.7.6, IPython 1.2.1, NumPy 1.8.2, Matplotlib 1.3.1
• Python 2.7.6, IPython 5.2.2, NumPy 1.12.0, Matplotlib 1.3.1
• Python 2.7.13, IPython 5.3.0, NumPy 1.11.3, Matplotlib 1.5.1
• Python 3.4.3, IPython 5.3.0, NumPy 1.14.5, Matplotlib 2.2.2
• Python 3.5.3, IPython 5.3.0, NumPy 1.12.1, Matplotlib 2.0.2
• Octave 4.0.2

Recent effort have also been conducted on an Ubuntu 16.04 x86_64 Linux system with the Python, NumPy, Matplotlib and Octave versions identified below.

• Python 2.7.12, IPython 2.4.1, NumPy 1.11.0, Matplotlib 1.5.1
• Python 3.5.2, IPython 7.1.1, NumPy 1.11.0, Matplotlib 3.0.2
• Octave 4.2.2

While Octave is not required to use RTB for Python, it is being utilized to run RTB for MATLAB^® demo and test scripts for comparison with updated RTB for Python demos and tests. Several MATLAB^® and Octave scripts in the RTB v9.8 distribution required modification to execute in Octave 4.0.2 and 4.2.2. The modified scripts are provided in the ./Octave/rvctools subdirectory of this distribution. As with the Octave scripts in the RTB distribution, simply copy the scripts to the corresponding ./rvctools subdirectories in the default or user specific Octave function search path.

Further Attention

• Non-convergent inverse kinematic iterations do not produce the same error norms among Python 2, 3 and Octave, which may indicate an implementation error in the ikine function or affects of possible differences in numerical precision of utilized linear algebra library functions. It was noted that on an Ubuntu 16.04 platform, Pythons 2.7.12 and 3.5.2 with Numpy 1.11.0 gave identical results for the inverse kinematic test cases which tends to indicate differing results may be attributed to the Numpy versions.
• Clamping joint movements in the ikine solver iteration loop is needed to produce realistic manipulator motion.
• RTB for Python and Octave 4 currently use 'lsode' method to integrate differential equations of motion for manipulator dynamics, but ode45 is available for both and could be a user selectable feature.
• Encountered several instances where fundamental differences between MATLAB^® and NumPy array/matrix operations (refer here), and Python type coercion and copy properties caused program errors in RTB for Python. Unidentified instances may still exist.
• Enhancements to robot pose graphics capabilities and features will require more sophisticated data structures for the Robot class (i.e., for the handle and plotopt attributes) and the introduction of a Plot class to eliminate plot module global variables and address redrawing event handling issues.
• RTB for Python module import statements require refinement to restrict overly broad inclusions of non-essential module components and to address circular references. The reliance on 'from module import *' statements hinders the ability for epydoc to produce only RTB for Python robot module specific documentation.
• Organization of modules into an object-oriented design pattern would improve the usability and maintainability of RTB for Python source code.
• Testing currently requires visual inspection of print and graphic test results to determine pass or fail outcome. A rigorous, automated testing strategy should be employed.

Attributions

• Luis Fernando Lara Tobar and Peter Corke for Robotics Toolbox for Python provided as the robotics-toolbox-python project.
• Yeison Cardona for setup script and Python 3 enhancements to the Robotics Toolbox for Python provided in the PyPi robotics-toolbox 0.1 project.
• Gary Deschaines for plot module 2D and 3D animation routines derived from the IK_Solver program.

Disclaimers

• See the file DISCLAIMER-RTB
• See the file DISCLAIMER-GED

Epilogue

A prime consideration in the RTB for Python update effort involves working with 10 year old Python code and 7 year old Octave scripts. Their relevance in the present time must be measured against the newest features and capabilities of RTB v10, Octave v4.4, Python v3.6, and the availability of numeric and graphic packages and libraries on commonly used computer platforms. Also, there is a modern predilection for realistic animation imagery and sophisticated GUI's that puts a significant software requirements burden on future development efforts with RTB for Python.

In order to concentrate on the correctness and accuracy of serial-chain kinematics and dynamics embodied in RTB for Python and Octave, less importance was placed on enhancing the user interface. To this end, the data plots and simplistic stick figure graphics were deemed sufficient to visualize important concepts, and to evaluate simulated manipulator kinematic and dynamic behavior. After the core functionality of RTB for Python has been updated, then user interface enhancements to provide the tutorial features and API available in newer versions of the Robotics Toolbox for MATLAB^® can be addressed.

Baseline Sources

• robotics-toolbox-python-master on June 23, 2018 from https://github.com/petercorke/robotics-toolbox-python
• robotics-toolbox-python v 1.8 on June 23, 2018 from https://code.google.com/archive/p/robotics-toolbox-python/downloads
• yeisoneng-robotics-toolbox-python-7aa5be5645f1 on June 23, 2018 from https://bitbucket.org/yeisoneng/robotics-toolbox-python
• robotics-toolbox-matlab-master on June 23, 2018 from https://github.com/petercorke/robotics-toolbox-matlab
• Robotics Toolbox v 9.8 on June 23, 2018 from http://www.petercorke.com/RTB/