Installation

1. SUMO

   Install SUMO using this link: https://sumo.dlr.de/docs/Installing/index.html

   For easy installation of dependencies and running the reinforcement learning model, we recommend installing SUMO in a Windows environment using the conda environment.

2. Anaconda

   Install Anaconda using this link: https://docs.anaconda.com/free/anaconda/install/windows/

After Successful installation of both SUMO and Anaconda/Miniconda, download the folder and create a conda environment using the enviornment.yaml file in the folder using the command given below:

cd <folder_location\\Sumo_AV_Gym>
conda env create -f environment.yaml

• Changes to be made in Sumo.py file
  • cd gym_sumo\envs
  • locate and open Sumo.py
  • Change sumoBinary location inside the initSimulator function in both if and else statements (# comments will be provided in the code to make sure you locate the code line)

After creating the conda environment with the required dependencies using the file, to activate the environment use the command given below:

conda activate sumo_av_gym

Training Phase

To train the autonomous vehicle using the provided environments, run the following command on your terminal

• Make sure to be inside the base folder which should be Sumo_AV_Gym

If you want to see the sumo environment gui while training, change the self.withGUI = False to self.withGUI = True inside the SUMOEnv class -> _init_ function -> mode="train" in the sumo_env.py file

If you want to start the training from the start without continuing from the previous model trained use the below command

python train.py

If you want to continue the training use the below command, the last model saved was around 1200 episdoes, so we need to increase the number of episodes to continue the training

python train.py --continue_train --no_of_episodes 1400

It is suggeested to run the training without the GUI to speed up the training process

Testing Phase

To test the pretrained rl model (which are already provided within the folder inside the models folder), run the following command on your terminal

• Make sure to be inside the base folder which should be Sumo_AV_Gym

python test.py --model 1200_episode_model.pth

You are free to choose any model you want to test which is in the models folder but recommended to test using the 1200_episode_model.pth