code up to this point coming from :
additional comments have been put and code separated into multiple classes
Before being able to go further and run the code, it is necessary that you download everything needed to run carla https://carla.readthedocs.io/en/latest/start_quickstart/
Secondly, you need to download python 3.7 as Carla latest version for windows is only available for python 3.7.
You may follow the videos 1 and 2 of the first link to get started as they show you how to start the simulation, how to spawn cars and even ride your own car in the environment.
When you have got acquanted with it, follow the third video that explains you how to spawn your own car (make it go straight for 5 seconds and then dissapear. When you are done, you will be able to follow the reinforcement learning.
As said before, I have already written all the code that has been written during the videos, however I have separated them to have a better visibility.
We have 3 classes and 1 script to view our model.
- Car environment
- DQN agent
- self made tensorboard
To run the code, first you need to place this folder where you have installed your carla application. More precisely, in the folder PythonAPI, same folder as carla's examples.
Next, you need to install the needed libraries to run your code, especially the following
- tqdm==4.43
- tensorflow==1.14 (not higher otherwise will not work, or you will have to change multiple things)
- Keras==2.2.5
To install them on your python 3.7, you can use the following command
py -3.7 -m pip install Keras=2.2.5Next, open a command line and go into the folder you just added with the different scipts (an easy way to that is by going into the folder in file explorer and type in navigation bar cmd.
As you are in the command line type the following
py -3.7 car_environment.pyif everything goes well, you should see a progress bar after a few seconds depending on your machine show the progress on the episodes of your machine learning.
finally, when the learning is done, you can view your model graphics with tensorflow.
To this, first you will need to add the tensorflow application to your path. Next, you can type the following command to open tensorflow web server
tensorboard --logdir==logs/If everything goes well, a message should appear telling you where you can go to view you tensorboard.
On your tensorboard you will be able to see something like this :
This can be done with the last script named model_viewer
Again to run it, same as before. An image viewer will then open with the view of the camera on the car. Before you can run it, you will need to specify the path to model that has been created during you car environment run (thus under the folder /model)