Autonomous Navigation in Complex Environments with Deep Multimodal Fusion Network - IROS 2020

System and softwares information

• Ubuntu 16.04/18.04

• ROS Kinetic/Medolic

• Gazebo 7/8

How to load the world files

• Download the 3D models (3GB), then extract and copy them to your Gazebo folder located at /home/<use_name>/.gazebo/models.

• Copy world files inside world_files folder into your catkin folder.

• The example world file is in navigation/simulation-env/world_files folder.

• Example command to run: rosrun gazebo_ros gazebo world_files/house4.world.

How to generate domain randomisation with random texture

• Make sure you pull all models from this git and copy them into your Gazebo model folder (which is located in /.gazebo/models)

• Make sure you had backed up your orginal models in ./gazebo/models. This is important since the code will delete your original textures and replace them. You will lose your original textures if you don't back up them!

• Open GENERATE.py and edit the path into your path of Gazebo models folder.

• Run the GENERATE.py file: python GENERATE.py This will automatically replace all old textures by new textures

• Reload your world files. You should see your worlds are in new colors.

• Known issues: Gazebo may immediately crash when loading a model or a world file. Fix this with: sudo apt-get install libgazebo7-dev

Build BeetleBot

$ cd navigation/beetlebot

$ catkin_make

$ . devel/setup.bash

$ roslaunch beetlebot_gazebo beetlebot_world.launch

(Wait until the world finishes loading in the gazebo)

$ roslaunch beetlebot_gazebo load_robot.launch

Control robot with keyboard:

$ cd navigation/beetlebot
$ . devel/setup.bash
$ cd src/beetlebot_AIOZ/beetlebot_control/scripts
$ python keyboard_controller.py

key	command
w	move forward
a	turn left
d	turn right
s	move backward
space	stop
r	rotate delivery-robot clockwise
f	rotate delivery-robot counter-clockwise
i	increase velocity by 1
k	decrease velocity by 1
j	decrease turning radius by 10 or increase turning degree
l	increase turning radius by 10 or decrease turning degree

Dataset

• Download the navigation dataset from here (58GB)

Training

To be update

Citation

If you use this code as part of any published research, we'd really appreciate it if you could cite the following paper:

@article{IROS2020aioz,
  author    = {Anh Nguyen and
               Ngoc Nguyen and
               Kim Tran and
               Erman Tjiputra and
               Quang D. Tran},
  title     = {Autonomous Navigation in Complex Environments with Deep Multimodal
               Fusion Network},
  journal   = {CoRR},
  volume    = {abs/2007.15945},
  year      = {2020}
}

@article{ROSBook2021aioz,
author = {Nguyen, Anh and Tran, Quang},
year = {2020},
month = {12},
journal = {ROSBook 2021 - Springer Book on Robot Operating System (ROS)},
title = {Autonomous Navigation with Mobile Robots using Deep Learning and the Robot Operating System}
}

License

AIOZ License

More information

AIOZ AI Homepage: https://ai.aioz.io AIOZ BeetleBot: https://beetle.aioz.io