Embedded Files
Hardware and parts
Hardware and parts
- Baxter robot (Rethink Robotics): https://www.rethinkrobotics.com/baxter/
- Electric parallel gripper: https://www.generationrobots.com/en/401517-electric-parallel-gripper-kit-for-baxter-research-robot.html
- Custom finger: made from the hilt of an Xacto knife inserted into a connector for the vacuum grippers
- Left and right arm cameras (see photo)
- Soft roll-up piano: https://www.amazon.com/Portable-Practice-Electronic-Microphone-Instrument/dp/B079J5K6XY/ref=sr_1_22_sspa?ie=UTF8&qid=1544746190&sr=8-22-spons&keywords=roll+up+piano&psc=1
- AR tags compatible with ar_track_alvar: http://wiki.ros.org/ar_track_alvar#ar_track_alvar.2BAC8-post-fuerte.Generating_AR_tags
Block diagram of code
Block diagram of code
Description of files:
Description of files:
Package: planning
Scripts:
demo_path.py : this script does everything – it computes the key poses, performs path planning, and plays the song.
piano.py: this script was not used in our final demo, but implemented the computer vision algorithm described here.
Launch files:
Src/planning/launch/baxter_grippers.launch: launch file for the MoveIt! Node, taken from a lab assignment
Src/ar_track_alvar/launch/ARTag_Resources/baxter_right_hand_track.launch and baxter_left_hand_track.launch: launch file for ar_track_alvar nodes. These include the intrinsic parameters for the left and right arm cameras and the physical size of the AR tag. This information is necessary for the homography computed internally in this package.
Understanding and using the system: a step-by-step guide
Understanding and using the system: a step-by-step guide
Run /run_demo.sh, a shell script for setting up the robot and launching ROS nodes
- Enable robot, open left and right arm cameras
- Launch ar_track_alvar for both left and right arm cameras
- Launch joint_trajectory_action_server
- Launch MoveIt! Node
- Run rviz to visualize tf frames on AR tags
Run /src/planning/src/demo_path.py, the main script
- Call make_note_dict(): gets transform from base frame to the two AR tag frames
- Call make_key_coords_reversed(): compute the poses of each key in the base frame. To do, we first carefully measured to find the poses of each key in the AR tag frames (the keys were just offset along the x-axis of these frames). Then, we just transformed the coordinates from the AR tag frames to the base frame using the rigid transformation found earlier.
- Call graph_markers() for debugging: publish markers at the note positions; RVIZ overlays them onto the camera feed to allow us to verify that we computed note positions correctly. See image on the left.
- Continue in main(): begins to play user-specified input. The user defines a list of melody notes and corresponding chords.
- Call play_melody_and_chords(): path plan to notes and execute, using MoveIt! Library functionalities. First, we insert waypoints (see image on left) between the notes to create an up-down pressing motion. Then, we set the orientation for each of the desired positions to be appendages pointing down. We also move the grippers apart or closer together to play the chord. Finally, MoveIt! Allows us to plan and execute for two arms simultaneously, which gives us the coordinated motion we want.
Report abuse