DriveLLM, The offical code repo for DriveLLM

This repository contains the official implementation of the DriveLLM system, which leverages large language model (LLM) capabilities to enhance autonomous driving decision-making processes. The LLM-service backend is implemented using Fast-API and langchain. The LLM-Autonomous driving bridge is implemented for ROS 1 designed for a modified version of Autoware stack

Other resources:

Testing rosbags recorded on all-weather autonomous shuttle bus (WATonoBus) project

instruction-following Finetune Dataset for General Driving Application

This contained a 20K dataset generated using self-instruct and is designed for general driving applications. Relevant dataset code on dataset generation, cleaning and formatting is here

Finetuned LLaMA model weight

• 13B LLaMA weight
• 7B LLaMA weight

How to use/test this service

Note: change http://localhost:9000/query to http://localhost:8300/query If you are running the service in docker.

Test the query endpoint:

Sending requests to the /query endpoint. Here is an example of using curl for service debuging (running locally):

curl -N -X POST 'http://localhost:9000/query' \
-H 'Content-Type: application/json' \
-d '{
    "messages": [
        {
            "role": "user",
            "perception":"perception", 
            "system_health":"system_health", 
            "weather":"weather", 
            "location":"location", 
            "vehicle_state":"vehicle_state",
            "control_command":"control_command",
            "command":"passenger_command"
        }
    ]
}'

🚀 Running the app in docker (recommended for deployment)

docker-compose up

Build the docker image and run the container.

docker-compose build
docker-compose up

If you need to completely shut down your environment or clean up your resources:

docker-compose down 

stops running containers and remove them, along with their associated networks, volumes, and images.

Note: For docker, the service will now be accessible at http://localhost:8300.

✅ Running locally (recommended for developer)

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

3. Create an .env file and add the following environment variables:

LOGGING_LEVEL=10 # 10-DEBUG, 20-INFO, 30-WARN, 40-ERROR

5. Run the application using Uvicorn:

uvicorn main:app --host 0.0.0.0 --port 9000

6. The service will now be accessible at http://localhost:9000.