SoLo T-DIRL

<<<<<<< HEAD |Paper | Video|

|Paper | Video|

c7eb0251fd6fe860040dbe37a65604f6dfc80fdf

This repository contains the codes for our ICRA2023 paper. For more details, please refer to the paper SoLo T-DIRL: Socially-Aware Dynamic Local Planner based on Trajectory-Ranked Deep Inverse Reinforcement Learning.

Abstract

This work proposes a new framework for a socially-aware dynamic local planner in crowded environments by building on the recently proposed Trajectory-ranked Maximum Entropy Deep Inverse Reinforcement Learning (T-MEDIRL). To address the social navigation problem, our multi-modal learning planner explicitly considers social interaction factors, as well as social-awareness factors into T-MEDIRL pipeline to learn a reward function from human demonstrations. Moreover, we propose a novel trajectory ranking score using the sudden velocity change of pedestrians around the robot to address the sub-optimality in human demonstrations. Our evaluation shows that this method can successfully make a robot navigate in a crowded social environment and outperforms the state-of-art social navigation methods in terms of the success rate, navigation time, and invasion rate.

Method Overview

Setup

1. Install tensorflow-1.12.0 gpu version.
2. Install Fetch gazebo simulation package.
3. Install pedsim simulation environment following the instruction.
4. Install PID controller ROS package inside your catkin workspace

git clone -b controller https://github.com/UMich-CURLY/Fetch_IRL.git tracking_pid

5. Install our SoLo TDIRL package

git clone https://github.com/UMich-CURLY/Fetch_IRL.git

6. Compile all packages.

cd ~/catkin_ws
catkin_make

Getting Started

This repository is organized into offline training and online testing.

Train a reward model.

1. Launch pedsim gazebo simulation environment.

roslaunch fetch_irl launch_dynamic.launch

2. Collect data. (Before collecting please change the data path and goal pose inside the main function of feature_expect.py)

roscd fetch_irl/script
python feature_expect.py
rosrun teleop_twist_keyboard teleop_twist_keyboard.py

3. Train the policy

python fetch_irl.py

Testing

1. Launch simulation environment.

roslaunch fetch_irl launch_dynamic.launch

2. Launch SoLo T-DIRL.

roscd fetch_irl/script
python main.py

3. Use 2D navigation goal button of RViz to naivgate the environment.

Simulation Videos

SoLo T-DIRL	SARL
LSTM-RL	CADRL

Citation

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

@misc{SoLo_TDIRL,
  doi = {10.48550/ARXIV.2209.07996},
  url = {https://arxiv.org/abs/2209.07996},
  author = {Xu, Yifan and Chakhachiro, Theodor and Kathuria, Tribhi and Ghaffari, Maani},
  keywords = {Robotics (cs.RO), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
  title = {SoLo T-DIRL: Socially-Aware Dynamic Local Planner based on Trajectory-Ranked Deep Inverse Reinforcement Learning},
  publisher = {arXiv},
  year = {2022},
  copyright = {Creative Commons Attribution 4.0 International}
}