In order to run you have first to train the cnn, we didn't push it to git since the file was over 500 MB.
To train the cnn you first need to generate the dataset. to do so launch gazebo with:
roslaunch project_assignment thymio_gazebo_bringup.launch
Then call the script from inside the scripts folder:
./create_dataset.py thymio10
Then press space to start generating the dataset.
Now you can train the cnn running from the scripts folder:
python3 main.py train
This will take a while. On a GTX 1070 it took around 30 min.
Now you can test the estimator running
roslaunch project_assignment thymio_gazebo_bringup.launch
./move.py thymio10
After launching press t to teleport the thymio to a known position.
Move using the keys:
w = move forward
x = move backward
s = stop moving
a = turn left
d = turn right
e = stop turning
t = teleport thymio to 0, 0.25, 0