- Requirements:
- Ubuntu 22.04 LTS
- docker
- python 3.6
- opencv-python
- psutil
- blosc
- pika
Python packages can be pip installed via requirements.txt.
The docker file must be built before running. This only needs to be done once! To do so run the following command:
sudo docker compose -f source/generator/portable-gui.yml build
The build process will take several minutes.
To run the visual demo enter the following command:
cd source/
python visual.py
- You will be prompted for a password as the python script internally calls docker compose.
- It can take up to 30 seconds for the window to appear.
- It can take up to 30 seconds for the demo to shutdown properly!
- You may see ~10 "run() Connection was closed, reconnecting...", this is normal.
--domain <val>{cartpole, vizdoom}, defaults to vizdoom.--difficulty <val>{easy, medium, hard} defaults to easy.--novelty <val>defaults to 200.
Example:
python visual.py --domain cartpole --novelty 201 --difficulty easy
To play select the OpenCV window and press the key corresponding to the desired action. One key press per action and frame update.
- Note: Two non-novel episodes will proceed the selected novel episodes.
- a → left
- d → right
- w → forward
- s→ backward
- j→ shoot
- k→ turn left
- l→ turn right
- q→ QUIT
- any other key is nothing
- a → left
- d → right
- w → forward
- s → backward
- q → QUIT
- any other key is nothing
- The doom game engine is a frame late for updating ammo used. Shoot commands act immediately and the ammo graphic will be updated next tick.
- If you cannot use docker on your system due to privilege issues, the dockerfile contains the installation instructions.
If the default port 55555 is already in use, another port can be specified with: --port <val>.
Similary with host: --host <val> like 127.0.0.10.
If you find a port communication bug try waiting for a minute.
If it is still an issue the docker might still be running. To kill it type:
docker ps
Find the ID of the latst container then enter:
docker kill <id>
If you see more than 20 of these you may need to change
docker compose --> docker-compose
inside of visual.py.