Welcome to L5Kit !¶

L5Kit is a Python library with functionality for the development and training of learned prediction, planning and simulation models for autonomous driving applications.

Warning

This project was discontinued and it is not actively maintained.

News¶

03-12-2022: We’ve just released a new notebook to train and evaluate the model introduced in our paper Safe Real-World Autonomous Driving by Learning to Predict and Plan with a Mixture of Experts !
10-06-2021: We’ve just released a series of notebooks to train and evaluate the model introduced in our paper Urban Driver: Learning to Drive from Real-world Demonstrations Using Policy Gradients !
08-20-2021: We’ve just released a new notebook tutorial describing our L5Kit gym-compatible environment!
04-16-2021: We’ve just released a new notebook tutorial for the ML simulation task!
12-03-2020: We’ve just released a series of notebooks to train and evaluate an ML planning model. We’ve also included pre-trained models!
11-26-2020: 2020 Kaggle Lyft Motion Prediction for Autonomous Vehicles Competition ended. We had more than 900 teams taking part in it!
11-16-2020: Dataset paper presented at CoRL 2020
09-29-2020: L5Kit v1.0.1 released
08-25-2020: 2020 Kaggle Lyft Motion Prediction for Autonomous Vehicles Competition started
08-24-2020: L5Kit v1.0.6 and Dataset v1.1 (includes traffic light support) released!