#ME495 Mini-Project: Sensable Phantom Omni and KDL

##Installing Phantom Omni Software and Drivers

To install the Phantom Omni Software on you computer follow these detailed instructions.

Exceptions that might occur:

1.After creating the udev rules for the Phantom Omni you will need to open the 50-phantom-firewire.rules file with sudo permissions to be able to edit and save it. The same applies for /etc/ld.so.conf.d which is created at a later stage. In example:

sudo gedit /lib/udev/rules.d/50-phantom-firewire.rules

2.After installtion, running

sudo dpkg -i openhaptics-ae_3.0-2_amd64.deb

from the OpenHaptics_Linux_v3_0/OpenHaptics-AE 3.0/64-bit folder, might not show the files libPHANToMIO.so or libPHANToMIO.so.4.3 as described in the instructions. In that case no duplicates were created, therefore one can skip the step of removing them from /usr/lib64. Your Omni Drivers will still work fine.

##Installing required ROS packages to interact with the Omni

You will need the following packages:

1. phantom_omni: Provides a node i.e. with rviz visualisation, joint state publishing, force manipulation and reporting button events.

2. omni_description:Contains the urdf model for the Sensable Phantom Omni

You can install these packages by cloning them into a correctly set-up [catkin workspace][1] from their respective github repositories. In the catkin_ws/src directory run: [1]:http://wiki.ros.org/catkin/Tutorials/create_a_workspace

git clone https://github.com/danepowell/phantom_omni
git clone https://github.com/danepowell/omni_description

Build your catkin workspace:

cd ~/catkin_ws
catkin_make

##Exploring phantom_omni and omni_description

From a terminal launch the phantom_omni package:

roslaunch phantom_omni omni.launch

Rviz should startup and a visualization of the omni will appear. Play around with the Omni!

In a new terminal explore the ROS topics:

rostopic list

In this project we will primarily be using the /omni1_joint_states and /omni1_force_feedback.

Get more information about a topic usingrostopic info /TOPICNAME i.e:

rostopic info /omni1_joint_states

The node /omni1 is publishing to /omni1_joint_states and omni1_robot_state_publisher is subscribing to it. The message data type is sensor_msgs/JointState.

To further see the description of the message type sensor_msgs/JointState run:

rosmsg show sensor_msgs/JointState

In this project we will be using the position component of JointState which can store the values for all the joint angles of the Omni.

If intersted in the data that is being published to a certain topic from the terminal, make use of: rostopic echo /TOPICNAME i.e:

rostopic echo /omni1_joint_states

Runnning the command view_frames listens to the /tf frames that are being broadcast and creates a frames.pdf file, which contains a tree of how the frames are conected.

##Creating additional packages

This section covers creating a new package. The nodes developed in this is package are capable of:

1.Using tf to lookup the transform from the frame located at the root of the Omni’s URDF to the frame at the end of the URDF

2.Using KDL to compute the forward, as well as inverse kinematics of the Omni

3.Using transformations.py to get the transforms from KDL and tf expressed as elements of SE(3)

4.Restricting the Phantom Omni to an axis floating in space

NOTE: If you do not want to manually recreate the demo package of this project, then clone me495_phantom_omni and build your catkin workspace analogous to the previous packages!!! (Skip to 'Exploring the ME495 Mini-Project)

###Create a package with the required dependencies

Following this tutorial, create a package in the catkin source directory:

cd ~/catkin_ws/src
catkin_create_pkg me495_phantom_omni omni_description phantom omni roscpp rospy tf

For our purposes you will depend on the following packages that need to be installed:

sudo apt-get install ros-indigo-robot-model
sudo apt-get install ros-indigo-urdfdom
sudo apt-get install ros-indigo-urdfdom-py

The above packages are debian packages that could be installed via apt-get. However, the hrl-kdl ROS package will also be depended on:

cd ~/catkin_ws/src
git clone https://github.com/gt-ros-pkg/hrl-kdl

###Writing the package node

Create a python script for your nodes:

gedit ~/catkin_ws/src/me495_phantom_omni/omni_mini_proj.py
gedit ~/catkin_ws/src/me495_phantom_omni/force_controller.py

Save them and make them executable

sudo chmod +x ~/catkin_ws/src/me495_phantom_omni/omni_mini_proj.py
sudo chmod +x ~/catkin_ws/src/me495_phantom_omni/force_controller.py

Paste the code from this repository into the respective python scripts.

Create a launchfile named omni.launch in me495_phantom_omni/src (same directory as the nodes) and paste this code into it.

Build your workspace with

catkin_make

You have now emulated cloning the repository and may move onto exploring the demo.

The package that has been cloned from github or recreated manually provides an elaborately documented code that implement the goals of this mini-project.

##Exploring the ME495 Mini-Project / me495_phantom_omni package

This project uses elementary P-conrollers to calculate constraint forces. The control optimization is not the focus of this project.

Caution: The Phantom's arm may go into resonance! GRIP THE STYLUS FIRMLY BEFORE STARTING THE LAUNCH FILE.

Launch the me495 mini-project demo by starting the launchfile. Grip the stylus firmly as it will jump to its constrained axis. The forces are set to be manageable. If it goes in resonance, simply grab the arm manually and stabilize it.

roslaunch me495_phantom_omni me495_omni.launch

As the nodes launch, rviz should begin to simulate the robot arm (this is a function of the phantom_omni package). The me495_phantom_omni package is publishing the desired constraint forces. It's also printing interesting feedback data in two separate terminals that should look like this:

[INSERT IMAGE]

The first matrix - an element of the SE(3) space - was calculated using the tf funcitons package. It calculates the transformation from the fixed frame located at the base of the urdf model /base to the body frame located near the tip of the stylus.

(Tip: use view_frames again to view the frames! Upon launching the project, the launchfile is publishing an additional fixed frame floating in mid-air, which is utilized for simplified constraint geometry calculations):

<node pkg="tf" type="static_transform_publisher" name="floating_frame_broadcaster"
	args="0.28 0 0.1 0 0 0  /base /floating 100" />

The second matrix being displayed in the large terminal is also an element of SE(3) that shows the same transformation as the previous one. It is obtained by parsing the urdf date to a kdl chain that captures the Omni's kinematics, subscribing to the Omni's joint states and finally computing the forward kinematics.

Using tf.transformtaions.is_same_transform (matrix1,matrix2) the two matrices are checked for equality.

To test the inverse kinematics, the joint angles q_sensors were offset offset by a constant to emulate an initial guess. Using the position of the stylus, the joint angles were obtained with the KDL inverse kinematics tools.

The bottom-most vector shows the difference delta_q between the sensor joint angles q_sensor and the ones obtained through inverse kinematics.

In a separate smaller window the constraint forces acting on the stylus are being displayed