Skip to main content

The oatomobile is a tool for developing and testing driving agents on the CARLA simulator

Project description

OATomobile: A research framework for autonomous driving

Overview | Installation | Baselines | Paper

PyPI Python Version PyPI version arXiv GitHub license

OATomobile is a library for autonomous driving research. OATomobile strives to expose simple, efficient, well-tuned and readable agents, that serve both as reference implementations of popular algorithms and as strong baselines, while still providing enough flexibility to do novel research.

Overview

If you just want to get started using OATomobile quickly, the first thing to know about the framework is that we wrap CARLA towns and scenarios in OpenAI gyms:

import oatomobile

# Initializes a CARLA environment.
environment = oatomobile.envs.CARLAEnv(town="Town01")

# Makes an initial observation.
observation = environment.reset()
done = False

while not done:
  # Selects a random action.
  action = environment.action_space.sample()
  observation, reward, done, info = environment.step(action)

  # Renders interactive display.
  environment.render(mode="human")

# Book-keeping: closes
environment.close()

Baselines can also be used out-of-the-box:

# Rule-based agents.
import oatomobile.baselines.rulebased

agent = oatomobile.baselines.rulebased.AutopilotAgent(environment)
action = agent.act(observation)

# Imitation-learners.
import torch
import oatomobile.baselines.torch

models = [oatomobile.baselines.torch.ImitativeModel() for _ in range(4)]
ckpts = ... # Paths to the model checkpoints.
for model, ckpt in zip(models, ckpts):
  model.load_state_dict(torch.load(ckpt))
agent = oatomobile.baselines.torch.RIPAgent(
  environment=environment,
  models=models,
  algorithm="WCM",
)
action = agent.act(observation)

Installation

We have tested OATomobile on Python 3.5.

  1. To install the core libraries (including CARLA, the backend simulator):

    # The path to download CARLA 0.9.6.
    export CARLA_ROOT=...
    mkdir -p $CARLA_ROOT
    
    # Downloads hosted binaries.
    wget http://carla-assets-internal.s3.amazonaws.com/Releases/Linux/CARLA_0.9.6.tar.gz
    
    # CARLA 0.9.6 installation.
    tar -xvzf CARLA_0.9.6.tar.gz -C $CARLA_ROOT
    
    # Installs CARLA 0.9.6 Python API.
    easy_install $CARLA_ROOT/PythonAPI/carla/dist/carla-0.9.6-py3.5-linux-x86_64.egg
    
  2. To install the OATomobile core API:

    pip install --upgrade pip setuptools
    pip install oatomobile
    
  3. To install dependencies for our PyTorch- or TensorFlow-based agents:

    pip install oatomobile[torch]
    # and/or
    pip install oatomobile[tf]
    

Citing OATomobile

If you use OATomobile in your work, please cite the accompanying technical report:

@inproceedings{filos2020can,
    title={Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts?},
    author={Filos, Angelos and
            Tigas, Panagiotis and
            McAllister, Rowan and
            Rhinehart, Nicholas and
            Levine, Sergey and
            Gal, Yarin},
    booktitle={International Conference on Machine Learning (ICML)},
    year={2020}
}

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for oatomobile, version 0.1.0
Filename, size File type Python version Upload date Hashes
Filename, size oatomobile-0.1.0.tar.gz (62.7 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page