##Systems Design Project - Group 7##

Rowan Border, Grant Kelly, Victor Dumitrescu, Ryan Davies, Ignotas Sulzenko, Martin Dimitrov, Maneshka Paiva, Mac Chong, Milan Pavlik

###Future SDP teams The planner and vision are very good.

We used an Arduino powered robot but the vision and planning can be effectively used even with the default NXT Brick. All it takes is to modify the Controller class in controller.py.

The vision is very accurate with the correct calibration and achieves around 28 FPS (faster than the camera). Planner is based on a reactive system. The defender strategy was one of the best in 2014 and the latest attacker strategy worked very well. In the end, it all depends on the build of the robot.

###Running the system

In the root of the project, execute python controller.py <pitch_number> <our_side> <our_color> where pitch_number is either 0 for the main pitch and 1 for the secondary pitch. Colors are regular yellow and blue. Side can be either left or right.

####Vision controls Press the following keys to switch the thresholding scheme: r - red p - plate b - black dots

Note that the colors of the plates themselves are ignored - you don't need them.

###Installation

To install the Polygon library, download the source, navigate inside and execute python setup.py install --user.

To isntall Argparse for python, download ArgParse, extract and run python setup.py install --user. All done.

There are a couple more libraries required: serial - can be removed if bluetooth is to be used (Use python-nxt to control the robot over bluetooth - bluez required)

Python-nxt requires bluez to be installed in order to communicate with the NXT over BlueTooth correctly. Assuming you are on a debian based Linux system, execute the following:

sudo apt-get install bluez.

To install the python library, navigate to nxt-python inside the lib directory and execute sudo python setup.py install to install the library globally. Alternatively, install virtualenv and do the following:

virtualenv venv
start ./venv/bin/activate
sudo python setup.py install

• At the moment OpenCV + Python are being used. A book on Computer Vision with OpenCV in Python is a decent starting point about what OpenCV can do.
• A detailed tutorial with examples and use cases can be found here - going through it can be handy to understand the code
• For OpenCV installation instructions please get in touch with others or have a look at the scripts in vision/

• Download OpenCV
• Extract the contents
• When on computers with with video feed, navigate to /disk/scratch/sdp/ and create a directory for OpenCV
• Copy extracted contents over to this directory
• Create directory build and navigate inside
• Execute

cmake -D CMAKE_INSTALL_PREFIX=~/.local ..` or *cmake -D WITH_OPENCL=OFF -D WITH_CUDA=OFF -D BUILD_opencv_gpu=OFF -D BUILD_opencv_gpuarithm=OFF -D BUILD_opencv_gpubgsegm=OFF -D BUILD_opencv_gpucodec=OFF -D BUILD_opencv_gpufeatures2d=OFF -D BUILD_opencv_gpufilters=OFF -D BUILD_opencv_gpuimgproc=OFF -D BUILD_opencv_gpulegacy=OFF -D BUILD_opencv_gpuoptflow=OFF -D BUILD_opencv_gpustereo=OFF -D BUILD_opencv_gpuwarping=OFF -D CMAKE_INSTALL_PREFIX=~/.local ..

to install without GPU libraries

• Execute make and wait (for quite some time)
• Execute make install
• Run ipython and do import cv2, if all executes fine then you're set.

The problem with installing OpenCV on OS X is because the Python library was installed in the wrong directory. To fix this..:

• Go to usr/local/share/bin
• Copy the contents from python2.7/site-packages to /Library/Python/2.7/site-packages