Skip to content

Homework for COMSW6731 (Spring 2019) Humanoid Robots at Columbia University

Notifications You must be signed in to change notification settings

jingxixu/humanoid-robots-s19

Repository files navigation

Homework for COMSW6731 Humanoid Robots (SP19) at Columbia University (Instructor: Prof. Peter Allen).

Table of Contents

Workspace Preparation

You need to have a computer with Ubuntu 14.04 and ROS Indigo installed. If you choose to use Ubuntu 16.04 or ROS Kinectic, it is your own responsibility to make sure that it also works on Ubuntu 14.04 + ROS Indigo, on which we tested your submission. The ROS Indigo installation link is here.

You might also want to install many dependencies if in need. Some helper commands can be found in clic_installation_14.04.sh and clic_installation_16.04.sh.

After you have successfully installed Ubuntu and ROS, you should run the following commands to create a workspace

mkdir -p ~/homework_ws/src
cd ~/homework_ws

Clone the following repositories into your workspace which we will use later.

cd ~/homework_ws/src
git clone https://github.com/jingxixu/fetch_gazebo.git
git clone https://github.com/cse481wi18/cse481wi18.git

You need to switch to branch gazebo2 of fetch_gazebo

cd ~/homework_ws/src/fetch_gazebo
git checkout gazebo2

Build this workspace and source it. Note, whenever you open a new terminal (unless you add this command to your .bashrc file), you need to source the setup.bash file.

cd ~/homework_ws
catkin_make
source devel/setup.bash

Part 1: Basic Motion (20%)

In this part you need to control the robot to do a series of basic motions. See a video demo here.

Prerequisites

We expect you to have already gone through the following tutorials before you start this part.

Launch

Run this command to launch a Fetch robot with a playgroud in Gazebo.

roslaunch fetch_gazebo playground.launch x:=0.0 y:=0.0 z:=0.0

Details

You are required to implement the following sequence of motions:

  • move base forward 1.5 meters
  • head looks down 45 degrees
  • head looks up 90 degrees
  • head looks left 90 degrees
  • head looks right 90 degrees
  • head looks forward
  • raise torso to the maximum
  • move arm to joint values [0, 0, 0, 0, 0, 0, 0]
  • close gripper
  • open gripper

You are allowed to use fetch_api package from the cse481wi18 repositiry that you cloned. You should name your response motion_demo.py and it is recommended that you put it under ~/homework_ws/src/fetch_gazebo/fetch_gazebo_demo/scripts/.

Part 2: Using MoveIt! (20%)

This part is quite easy! You just need to show a video of you using MoveIt! RViz plugin to sample a random valid goal joint values for the arm group, plan and then execute. Then use the MotionPlanning - Slider to replay the plan or go over the plan by waypoints. See a video demo here.

Prerequisites

We expect you to have already gone through the following tutorials before you start this part.

Launch

Launch a Fetch robot with a playgroud in Gazebo

roslaunch fetch_gazebo playground.launch x:=0.0 y:=0.0 z:=0.0

Launch MoveIt! for Fetch

roslaunch fetch_moveit_config move_group.launch

Launch RViz (you might need to add a RobotModel and MotionPlanning visualization using the add botton)

rosrun rviz rviz

Part 3: Using GraspIt! (20%)

In this part, you are required to create a scene in GraspIt! and generate grasps. See a video demo here.

Prerequisites

Follow the instructions here to install GraspIt! You should also go over those simple usages.

Details

The graspable object you are using is longBox and the gripper is fetch_gripper. You are required to include a obstacle floor and put the longBox on the floor as shown in the following image. This requires you to figure out a right pose to load longBox and floor in GraspIt!. GraspIt! has built-in functions to load these objects which allows you to specify their poses. Name your python file graspit_demo.py and put it under ~/homework_ws/src/fetch_gazebo/fetch_gazebo_demo/scripts/.

Part 4: Pick Demo (40%)

In this part you are required to make the robot pick the long box using the grasp generated by GraspIt!. See a video demo here.

Launch

Launch a Fetch robot with a playgroud in Gazebo

roslaunch fetch_gazebo playground.launch

This time we do not specify where to start the robot so it is now at a default position, as shown in the below image. The long box on the table is the same as the one you used in part 3 so that the grasps you generated will work.

Name your python file pick_demo.py and put it under ~/homework_ws/src/fetch_gazebo/fetch_gazebo_demo/scripts/. Run the following launch file to start your script as well as MoveIt!

roslaunch fetch_gazebo_demo pick_demo.launch

Details

You are required to make the robot:

  • go to the table
  • pick the object
  • lift it up
  • return to the original position

There are many ways to do it. The only requirement is that you should not use perception and you are required to use the grasp generated by GraspIt!.

Here are some hints. The grasps returned by GraspIt! is in the frame of the GraspIt! world. You need to transform them into using robot's base_link, so that you can call MoveIt inside your program to move the arm to that pose. You might want to look at Gazebo's gazebo_msgs/GetModelState service and the python package tf.transformations. You will also need to deal with collision while using MoveIt!, you need to add collision objects into the scene. An example of the created scene in MoveIt! is shown below. You do not need to create the exact scene as long as it works.

NOTE: It is understandable if your script does not work 100% of the times, as there are too many unpredictable factors. But you need to at least get it work in your video demo.

Submission

Submit a folder named Name_UNI, for example, john_white_jw2333. The folder should contain the following files:

  • motion_demo.py
    • your code response for part 1
  • graspit_demo.py
    • your code response for part 3
  • pick_demo.py
    • your code response for part 4
  • README.md
    • all 4 links of your YouTube video demo.
    • any other implementation details that you want we know
  • (optional) other scripts or files that your code depends on

Compress this folder to Name_UNI.zip and submit it to Coursework.

About

Homework for COMSW6731 (Spring 2019) Humanoid Robots at Columbia University

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages