Self-Learning Exploration and Mapping for Mobile Robots via Deep Reinforcement Learning

September 15, 2020 · View on GitHub

This repository contains code for robot exploration with Deep Reinforcement Learning (DRL). The agent utilizes the local structure of the environment to predict robot’s optimal sensing action. A demonstration video can be found here.

drawing

drawing

Dependency

  • Python 3
  • scikit-image 
    pip3 install scikit-image
  • tensorboardX 
    pip3 install tensorboardX
  • TensorFlow (this code is writen under TF1.x but it is modified to be compatible with TF2)
  • pybind11 (pybind11 — Seamless operability between C++11 and Python) 
    git clone https://github.com/pybind/pybind11.git
cd pybind11
mkdir build && cd build
cmake ..
sudo make install

Compile

You can use the following commands to download and compile the package.

git clone https://github.com/RobustFieldAutonomyLab/DRL_robot_exploration.git
cd DRL_robot_exploration
mkdir build && cd build
cmake ..
make

How to Run?

  • For the CNN policy:

    cd DRL_robot_exploration/scripts
python3 tf_policy_cnn.py

  • For the RNN policy:

    cd DRL_robot_exploration/scripts
python3 tf_policy_rnn.py

  • To select the running mode, at the beginning of the tf_policy code:

    # select mode
TRAIN = False
PLOT = True

    Set TRAIN=False to run the saved policy. You can train your own policy by setting TRAIN=True. Set PLOT=True to show visualization plots.

  • To show the average reward during the training:

    cd DRL_robot_exploration
tensorboard --logdir=log

Cite

Please cite our paper if you use any of this code:

@inproceedings{ExplorDRL2019,
  title={Self-Learning Exploration and Mapping for Mobile Robots via Deep Reinforcement Learning},
  author={Chen, Fanfei and Bai, Shi and Shan, Tixiao and Englot, Brendan},
  booktitle={AIAA SciTech Forum},
  pages={0396},
  year={2019},
}

Reference