Home

Awesome

Social-NCE + CrowdNav

Website | Paper | Video | Social NCE + Trajectron | Social NCE + STGCNN

<p align="center"> <img src="docs/illustration.png" width="300"> </p>

This is an official implementation for
Social NCE: Contrastive Learning of Socially-aware Motion Representations <br> <a href="https://sites.google.com/view/yuejiangliu/">Yuejiang Liu</a>, <a href="https://qiyan98.github.io/">Qi Yan</a>, <a href="https://people.epfl.ch/alexandre.alahi/?lang=en/">Alexandre Alahi</a>, ICCV 2021 <br>

TL;DR: Contrastive Representation Learning + Negative Data Augmentations 🡲 Robust Neural Motion Models

[New] our more recent work on this topic:
Towards Robust and Adaptive Motion Forecasting: A Causal Representation Perspective, CVPR 2022.

Preparation

Setup environments follwoing the SETUP.md

Training & Evaluation

Basic Results

Results of behavioral cloning with different methods.

<img src="docs/collision.png" height="240"/> <img src="docs/reward.png" height="240"/>

Averaged results from the 150th to 200th epochs.

<table><tbody> <!-- START TABLE --> <!-- TABLE HEADER --> <th valign="bottom"></th> <th valign="bottom">collision</th> <th valign="bottom">reward</th> <!-- TABLE BODY --> <tr><td align="left">Vanilla</td> <td align="center">12.7% &#177; 3.8%</td> <td align="center">0.274 &#177; 0.019</td> <tr><td align="left">Local</td> <td align="center">19.3% &#177; 4.2%</td> <td align="center">0.240 &#177; 0.021</td> <tr><td align="left">Ours</td> <td align="center">2.0% &#177; 0.6%</td> <td align="center">0.331 &#177; 0.003</td> </tr> </tbody></table>

Citation

If you find this code useful for your research, please cite our papers:

@inproceedings{liu2021social,
  title={Social nce: Contrastive learning of socially-aware motion representations},
  author={Liu, Yuejiang and Yan, Qi and Alahi, Alexandre},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  pages={15118--15129},
  year={2021}
}
@inproceedings{chen2019crowd,
  title={Crowd-robot interaction: Crowd-aware robot navigation with attention-based deep reinforcement learning},
  author={Chen, Changan and Liu, Yuejiang and Kreiss, Sven and Alahi, Alexandre},
  booktitle={International Conference on Robotics and Automation (ICRA)},
  pages={6015--6022},
  year={2019}
}