Tetris Player

Plan

Develop a piece-recognition node.
- This node will detect the shapes of a given set of Tetris pieces placed in a starting area using Baxter’s cameras.
Develop a Tetris solver.
- Given a set of Tetris pieces, the solver will return a complete placement for them in the form of an array representing the board. This will be our solution configuration.
Develop a piece-placing node.
- This node will use Baxter’s suction end effector to place the pieces in the solution configuration using closed-loop control.

Getting Started

First-time setup:

$ pip2 install --user -r requirements.txt
$ echo 'export PATH=$PATH:~/.local/bin' >> ~/.bashrc
$ echo 'export PYTHONPATH=$PYTHONPATH:~/.local/share/python2.7/site-packages' >> ~/.bashrc
$ source ~/.bashrc

For connecting to a robot (need to run every session):

$ export ROBOT_NAME=[name]
$ ./baxter.sh "$ROBOT_NAME".local
$ rosrun baxter_tools enable_robot.py -e

For pick-and-place

roslaunch baxter_moveit_gui_noexec.launch for RViz GUI path planning
./baxter.sh [name].local which is either asimov, ayrton, archytas, ada, or alan

Enable Baxter by running rosrun baxter_tools enable_robot.py -e
Start the Baxter trajectory controller by running rosrun baxter_interface joint_trajectory_action_server.py
Start MoveIt for Baxter by running roslaunch baxter_moveit_config demo_baxter.launch right_electric_gripper:=true left_electric_gripper:=true omitting the last argument if the robot lacks (a) gripper(s). If this fails on Baxter, try roslaunch baxter_moveit_config baxter_grippers.launch MoveIt is now ready to compute and execute trajectories on the robot

rosrun tetris path_test.py
rosrun tf tf_echo base reference/right_gripper
rosrun baxter_tools tuck_arms.py -u

Cameras

`head_camera', 'left_hand_camera', 'right_hand_camera'
rosrun baxter_tools camera_control.py -c head_camera
rosrun baxter_tools camera_control.py -o left_hand_camera -r 1280x800

TODO

Detect what pieces there are ?
Input the pieces into the solver, get solution !
From solution figure out location (coordinates) to place pieces !
Detect where to pick up the pieces for each piece
Place them in their correct positions (how to get accuracy right?)

Documentation

To view the project site locally, run:

$ cd docs
$ bundle install
$ bundle exec jekyll serve

Troubleshooting

Camera field-of-view is too close. Use resetcameras to set the resolution to 1280 by 800. Note that you may have to disable the right-hand camera to get a bandwidth boost.
Robot is dropping the first tile in the lower right-hand corner. The board is probably oriented incorrectly (tile 8 should be in the top left).
The AR tags are failing to be detected. You may have to increase lighting so the light parts of the tags register.
Suction is not working. First, ensure the pressurizer is on. Next, the seal on the gripper may be poor. You can check whether this is the case by running:
```
$ roslaunch tetris baxter_moveit_headless.launch
$ rosrun tetris gripper_test.py
```
Then, feel the seal around the tube while the gripper is activated. If large amounts of air are moving through the seal, the best fix is to put layers of tape between the tube screw and the socket (not around the seal). The likely cause of the failure is that the screw is bent such that it creates gaps, even when fully screwed in. The tape fills those gaps, since it is somewhat compressible.
Cannot find a path. Some points may not be feasible. Make sure your table is high enough, and fully within reach of the robot.

If all else fails, you can reboot the robot. This has been known to solve numerous issues, like being unable to plan paths that are clearly feasible.

xtyxtyx/tetris