Limited Offer Get 25% off — use code BESTW25
No AI No Plagiarism On-Time Delivery Free Revisions
Claim Now
Prices start at $5 per page · plagiarism-free papers · free revisions for 14 days 24/7 support: support@bestwriters.com
Uncategorized

COMP9069 Robotics & Autonomous Systems

  • Posted
  • Authoradmin
  • Length854 words
  • Reading time4 min
Write My Assignment FREE

COMP9069 Robotics & Autonomous Systems – Assignment 2
Due: Sun 10th May 23:59
Once completed, upload all of your code to Canvas in a single archive (zip, tar, etc.).
Your submission should include:
• All ROS packages created for this assignment (my_ur3_description and my_ur3_reach).
• All openai_ros files created or modified for this assignment.
Preliminaries:
If you have not done so already, install ROS effort controllers and Joint State Publisher GUI, and
configure Python3 support:
$ sudo apt-get install ros-melodic-joint-state-publisher-gui ros-melodic-effort-controllers
$ sudo apt-get install python3-pip python3-yaml
$ python3 -m pip install rospkg catkin_pkg
Additionally, install TensorFlow, OpenAI Gym and OpenAI Baslines for Python3 (see details in
Lab 08 and 10).

Task 1
Create a package named my_ur3_description with the following dependencies:
[10%]

• rviz
• controller_manager
• gazebo_ros
• joint_state_publisher
• robot_state_publisher
Create the following folders within the package directory:
• launch
• meshes
• models
• robots
Download the file Assignment2_Task1.zip from Canvas and extract. Place the contents of launch in
your launch folder, meshes in your meshes folder, etc.
Execute test_rviz.launch to view the robot in RVIZ and familiarise yourself with its structure and
workings.
Modify your package to add controllers to the UR3 robot. Use controller types
effort_controllers/JointPositionController.
Once finished, execute ur3.launch to start the simulation (note: you will need to edit one of the
launch files, as indicated in comments, to include your controller launch file).
Use the Dynamic Reconfigure tool within rqt_gui to find good PID gain values for the controllers.
Edit your controller configuration file to save these values.
Task 2 [20%]
Create a new package, my_ur3_reach. Within this package, create a ROS node that prompts the user
for a position to move a given joint to. The available joints can be presented to the user in a textbased menu. Once the user selects a given joint, they should be prompted for the position to set that
joint to. The node should then move the joint to the specified position and print:
• The final actual position of the joint.
• The error in the position, relative to the desired set-point.
• The position and orientation of the end effector relative to the base.
Task 3 [30%]
Create an openai_ros robot environment for your UR3 robot. Skeleton code is provided in
ur3_env.py, which is available on Canvas.
Additional notes:
• If working from CartPole example, be aware that there was only one joint to move for the
CartPole. Here, there are multiple joints to move.
• When moving the joints, you should wait for them to reach the desired positions before
proceeding with further execution. Note, however, that there will be some error in the final
positions.
Task 4 [30%]
Create an openai_ros task environment for the training of the UR3 robot. The task to be
accomplished is simply to move the robot arm joints (two shoulder joints and the elbow) to
specified positions. As such, the observation space is simply the positions of these three joints.
When the desired positions are achieved, the end effector should be approximately at the position of
the cube in the simulation.
The reward at each step is defined as follows:
r=
1
√∑
i
d
2i
where di is the difference between the current position of joint i and its desired setpoint. An
additional reward of 100 (as specified in the task configuration file) is obtained if the robot achieves
the desired pose, with an absolute tolerance of 0.2 per joint, i.e. if all arm joints are within 0.2 of
their desired setpoints.
There are six available actions corresponding to +/- movements of each arm joint.
An episode is complete (i.e. done) if the robot achieves the desired pose, with an absolute tolerance
of 0.2 per joint, or if the maximum number of iterations is exceeded.
The task configuration file is available on Canvas: ur3_position_task.yaml
Note: in addition to the task environment Python code, you will need to make further modifications
to the openai_ros package in order to be able to use this environment.
Task 5 [10%]
Train the UR3 robot to accomplish the task defined in task 4. Implement the following RL
algorithms:
• Q-learning.
• SARSA.
• DQN.
Notes:
The task to be solved here is trivial and obviously does not require an RL solution. It doesn’t even
require inverse kinematics! If we were unconcerned with the trajectory, we could simply send the
setpoints as position commands to appropriate controllers. There may be some error in the final
positions but we could minimise that by tuning the contoller PID gains.
Also, we haven’t considered any workspace collisions (e.g. collision between the robot arm and the
table) or the fact that the robot arm may end up pushing the cube off the table (we could introduce
penalties for these outcomes but it slightly complicates the task environment).
Lastly, we wouldn’t typically control the movement of the robot in the way described here
(independently moving each joint). Instead, we would typically use ROS MoveIt! for trajectory
planning. Although MoveIt! is relatively straightforward, this trajectory planning again complicates
the task and robot environments and is beyond the scope of this assignment.

The post COMP9069 Robotics & Autonomous Systems appeared first on My Assignment Online.

Plagiarism Free Assignment Help

Expert Help With This Assignment — On Your Terms

✓ Native UK, USA & Australia writers ✓ Deadline from 3 hours ✓ 100% Plagiarism-Free — Turnitin included ✓ Unlimited free revisions ✓ Free to submit — compare quotes

admin

Academic writer and editor at BestWriters.

Get your paper off your plate tonight

Post your instructions for free. Prices start at $5 per page.

Place an order