Skip to content

Latest commit

 

History

History
82 lines (56 loc) · 2.67 KB

quickstart.md

File metadata and controls

82 lines (56 loc) · 2.67 KB

TrafficManager Quickstart

💡 Please run the following code on a local machine with a display.

1. Install Python dependencies:

cd TrafficManager
conda create --name drivearena_tf python=3.8
conda activate drivearena_tf
pip install -r requirements.txt

After completing the setup, you can perform an initial verification:

cd ../ # under DriveArena root folder
python ./TrafficManager/sim_manager_only.py

If the program runs correctly and the GUI displays properly, proceed to the next step.

2. Deploy World Dreamer and Driving Agent Services on the remote Server or Locally

3. Port Mapping

💡 If WorldDreamer and Driving Agent are running locally or can be accessed directly via the public internet, you can skip this step.

Map the server-side service ports to your local machine:

# Assume you can connect to server as: ssh username@server_adress -p server_port
ssh -N username@server_adress -p server_port -L 11000:localhost:server_diffusion_port -L 11001:localhost:server_driver_port

Please keep this terminal window open.

4. Configuration Settings

Confirm that the IP and port settings are correct in the config.yaml file:

servers:
    diffusion: "http://127.0.0.1:11000/"
    driver: "http://127.0.0.1:11001/"

5. Run the Program

# under DriveArena root folder

# UniAD
python ./TrafficManager/sim_manager.py

# VAD
python ./TrafficManager/sim_manage_vad.py

The program should display the following interface:

Simulation Interface of TrafficManager

The program will stop either after the ego_agent completes the assigned route or if a collision occurs. You can also manually terminate the program at any time.

6. View Results The result files from each run are saved in subfolders within the DriveArena/results/ directory, named according to the run time in the format mm-dd-hhmmss.

  • Calculate the Driving Score:

    python ./TrafficManager/score_calculator.py ./results/mm-dd-hhmmss/ # modify to real path

  • Create a Video: The imgs/ subfolder inside the results folder contains images of each frame, including WorldDreamer outputs, GT BEV, and Agent-predicted BEV. Use the following script to make a video from these images:

    python ./TrafficManager/generate_video.py --output_dir   ./results/mm-dd-hhmmss/ # modify to real path

    A file named output_video.mp4 will be generated in the mm-dd-hhmmss folder.