Safe Real-World Autonomous Driving by Learning to Predict and Plan with a Mixture of Experts

Stefano Pini1
Christian S. Perone1
Aayush Ahuja2
Ana Sofia Rufino Ferreira2
Moritz Niendorf2
Sergey Zagoruyko1
1Woven Planet United Kingdom Limited
2Woven Planet North America, Inc
NeurIPS 2022 workshop on Machine Learning for Autonomous Driving (ML4AD)
High-level overview of SafePathNet, a ML approach improving on-road safety of self-driving vehicles (SDVs).
High-level overview of SafePathNet, a ML approach improving on-road safety of self-driving vehicles (SDVs).

Abstract

The goal of autonomous vehicles is to navigate public roads safely and comfortably. To enforce safety, traditional planning approaches rely on handcrafted rules to generate trajectories. Machine learning-based systems, on the other hand, scale with data and are able to learn more complex behaviors. However, they often ignore that agents and self-driving vehicle trajectory distributions can be leveraged to improve safety. In this paper, we propose modeling a distribution over multiple future trajectories for both the self-driving vehicle and other road agents, using a unified neural network architecture for prediction and planning. During inference, we select the planning trajectory that minimizes a cost taking into account safety and the predicted probabilities. Our approach does not depend on any rule-based planners for trajectory generation or optimization, improves with more training data and is simple to implement. We extensively evaluate our method through a realistic simulator and show that the predicted trajectory distribution corresponds to different driving profiles. We also successfully deploy it on a self-driving vehicle on urban public roads, confirming that it drives safely without compromising comfort. The code for training and testing our model on a public prediction dataset and the video of the road test are available at https://woven.mobi/safepathnet.

Road test

We evaluated our approach through a realistic simulator and tested it on a real SDV in both our private testing facility and on public roads. Results show that SafePathNet presents a better trade-off between comfort (discomfort braking, passiveness) and safety (collisions) and brings us another step closer to our goal of safe real-world autonomous driving.

Citation

If you find our paper or code useful, please cite our work as:

@article{pini2022safe,
  title={Safe Real-World Autonomous Driving by Learning to Predict and Plan with a Mixture of Experts},
  author={Pini, Stefano and Perone, Christian S and Ahuja, Aayush and Ferreira, Ana Sofia Rufino and Niendorf, Moritz and Zagoruyko, Sergey},
  journal={arXiv preprint arXiv:2211.02131},
  year={2022}
}