Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.


This repository contains the codes for our ICRA 2018 submission. For more details, please refer to the paper Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning.

If you find the codes or paper useful for your research, please cite our paper:

Author = {Changan Chen and Yuejiang Liu and Sven Kreiss and Alexandre Alahi},
Title = {Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning},
Year = {2018},
Eprint = {arXiv:1809.08835},


Mobility in an effective and socially-compliant manner is an essential yet challenging task for robots operating in crowded spaces. Recent works have shown the power of deep reinforcement learning techniques to learn socially cooperative policies. However, their cooperation ability deteriorates as the crowd grows since they typically relax the problem as a one-way Human-Robot interaction problem. In this work, we want to go beyond first-order Human-Robot interaction and more explicitly model Crowd-Robot Interaction (CRI). We propose to (i) rethink pairwise interactions with a self-attention mechanism, and (ii) jointly model Human-Robot as well as Human-Human interactions in the deep reinforcement learning framework. Our model captures the Human-Human interactions occurring in dense crowds that indirectly affects the robot's anticipation capability. Our proposed attentive pooling mechanism learns the collective importance of neighboring humans with respect to their future states. Various experiments demonstrate that our model can anticipate human dynamics and navigate in crowds with time efficiency, outperforming state-of-the-art methods.

Method Overview


  1. Install Python-RVO2 library
  2. Install crowd_sim and crowd_nav into pip
pip install -e .

Getting started

This repository are organized in two parts: gym_crowd/ folder contains the simulation environment and crowd_nav/ folder contains codes for training and testing the policies. Details of the simulation framework can be found here. Below are the instructions for training and testing policies, and they should be executed inside the crowd_nav/ folder.

  1. Train a policy.
python --policy sarl
  1. Test policies with 500 test cases.
python --policy orca --phase test
python --policy sarl --model_dir data/output --phase test
  1. Run policy for one episode and visualize the result.
python --policy orca --phase test --visualize --test_case 0
python --policy sarl --model_dir data/output --phase test --visualize --test_case 0
  1. Plot training curve
python utils/ data/output/output.log

Simulation Videos


Learning Curve

Learning curve comparison between different methods in invisible setting.