Carla Reinforcement Learning

This repository tracks the tests to use reinforcement learning in carla.

Preperation

• Download Carla0.9.13
• use pip carla api

Create the reinforcement learning python environment

conda env create -f environment.yml

conda activate carla-rl

conda activate myenv
conda env update --file local.yml --prune

carla

0.9.10.1 requires -opengl during the server startup due to issue:

./CarlaUE4.sh -opengl

0.9.13 works fine

check your python path

changing pythonpath requires restart? set pythonpath in ~/.bashrc

export PYTHONPATH=$PYTHONPATH:/home/luttkule/git/carla0.9.10.1-bin/PythonAPI/carla/dist/carla-0.9.10-py3.7-linux-x86_64.egg

check pythonpath with

python -c "import carla; print(carla.__file__)"

change map

https://github.com/carla-simulator/carla/issues/4901

CarlaUE4/Config/DefaultEngine.ini

not feasible to change default map easily

conda

conda env create -f environment.yml

currently conda activate carla-rl-py37-c13

IP address

provide the IP adress of the carla server in the ip-host.txt file if you want to run the server somewhere else and use the carla_env.read_IP_from_file() function

used https://stackoverflow.com/questions/4348590/how-can-i-make-git-ignore-future-revisions-to-a-file to ignore future changes to the file

git update-index --skip-worktree ip-host.txt