Course material for the Multi-agents and computer graphics course

Nov 27, 2021 4 min read

TC2008B

Course material for the Multi-agents and computer graphics course.

Setup instructions

Strongly recommend using a custom conda environment.
Install python 3.8 in the environment: conda install python=3.8 Using 3.8 for compatibility reasons. Maybe 3.9 or 3.10 are compatible with all the packages, but will have to check.
Installing mesa: pip install mesa
Installing flask to mount the service: pip install flask
By this moment, the environment will have all the packages needed for the project to run.

Instructions to run the local server and the Unity application

Run either the python web server: Server/tc2008B_server.py, or the flask server: Server/tc2008B_flask.py. Flask is considerably easier to setup and use, and I strongly recommend its use over python’s http.server module. Additionally, IBM cloud example used flask.
To run the python web server:

python tc2008B_server.py

To run a flask app:

export FLASK_APP=tc_2008B_flash.py
flask run

You can change the name of the app you want to run by changing the environment variable FLASK_APP.
Alternatively, if you used the following code in your flask server:

if __name__=='__main__':
    app.run(host="localhost", port=8585, debug=True)

you can run it using:

python tc2008B_flask.py

To run a flask app on a different host or port:

flask run --host=0.0.0.0 --port=8585

Either of these servers is what will run on the cloud.
Once the server is running, launch the Unity scene TC2008B that is in the folder: IntegrationTest.
The scene has two game objects: AgentController and AgentControllerUpdate. I left both so that different functionality can be tested: AgentController works with the response of the python web server, while AgentControllerUpdate works with the reponse from the flask server.
I updated the AgentController.cs code, and introduced AgentControllerUpdate.cs. Each script parses data differently, depending on the response from either the python web server, or from the flask server. The AgentController.cs script parses text data, while AgentControllerUpdate.cs parses JSON data. I strongly recommend that we use JSON data.
The scripts are listening to port 8585 (http://localhost:8585). Double check that your server is launching on that port; specially if you are using a flask server.
If the Unity application is not running, or has import issues, I included the Unity package that has the scene Sergio Ruiz provided.

Instruction to run the cloud server and Unity application

Installing dependencies, and locally running the sample

Following instructions from the IBM-Cloud repo
Download and install Cloud Foundry CLI. I am installing it on my WSL machine:

# ...first add the Cloud Foundry Foundation public key and package repository to your system
wget -q -O - https://packages.cloudfoundry.org/debian/cli.cloudfoundry.org.key | sudo apt-key add -
echo "deb https://packages.cloudfoundry.org/debian stable main" | sudo tee /etc/apt/sources.list.d/cloudfoundry-cli.list
# ...then, update your local package index, then finally install the cf CLI
sudo apt update
sudo apt install cf8-cli

To get the sample app running:

git clone https://github.com/IBM-Cloud/get-started-python
cd get-started-python

To run locally:

pip install -r requirements.txt
python hello.py

To deply the sample to the cloud

All the requiered files for the sample app to run are inside the IBMCloud folder.
We first need a manifest.yml file. The one provided in the example repository contains the following:

applications:
 - name: GetStartedPython
   random-route: true
   memory: 128M

You can use the Cloud Foundry CLI to deploy apps. Choose your API endpoint:

cf api

Replace the API-endpoint in the command with an API endpoint from the following list:

URL	Region
https://api.ng.bluemix.net	US South
https://api.eu-de.bluemix.net	Germany
https://api.eu-gb.bluemix.net	United Kingdom
https://api.au-syd.bluemix.net	Sydney

cf login

From within the get-started-python directory push your app to IBM Cloud:

cf push

This process can take a while. All the dependencies are downloaded and installed, and the app in started.
After you push the application, in the cloud dashboard you can see a new cloud foundry app.
This can take a minute. If there is an error in the deployment process you can use the command cf logs
--recent to troubleshoot.
When deployment completes you should see a message indicating that your app is running. View your app at the URL listed in the output of the push command. You can also issue the cf apps.
With the cf apps command you can see the route for the app.

To deploy a custom app to the cloud

I created an app within the cloud foundry in the ibm cloud by following the document Manual IBM Cloud - Python.pdf.
Created an additional folder inside the IBMCloud folder, named boids, that contains the required files.
In the manifest.yml I renamed the name to the one I used for the app in cloud foundry. From GetStartedPython to Boids.
Then, modified the ProcFile file as follows:

web: python tc2008B_flask.py

Modified the setup.py file, but I do not think it matters.
Then changed to the boids folder, and used:

cf push

Then, update the url for the service in Unity with the url for the service that cloud foundry assigns.

Notes

Using VSCode to develop everything.
Although not stated in the requirements, Git needs to be installed on the system.
I am running windows, and using the WSL. I ran the server code in WSL, and the Unity client in windows. My WSL machine runs Ubuntu 20.
Using Thunder Client extension as a replacement for postman to test the apis.
Pip does not allow us to search anymore.
As of 2021-10-17, the WWWForm method to post from Unity to the web service still works with Unity 20.20.3.4. However, the support apparently is going away soon.
Using flask because it is ideal for building smaller applications. Django could be used, but since it is much more robust, the additional utilities were not needed for this project.
The demo app push process went rather smoothly, but for the boids app it did not. It took too long, and ended up failing with a timeout error. I issued the command again.
Timeout again. Modified the manifest, and tried again.
After that, the app failed when it tried to start. Apparently, numpy was missing from the requirements.