Indoor Mobile Robot Fully Autonomous Driving Project - DPoom 
For further information, please visit our page
Click below for introduction video.
Single board computer Lattepanda Alpha based indoor mobile robot DPoom, with fully autonomous driving system using single RGB-D camera Intel Realsense D435i.
Keywords: autonomy, autonomous driving system, mobile robot, SLAM, ROS, RGB-D, Low-end, global path planning, motion planning, ground segmentaion, navigation, path tracking, control, Human-Robot Interaction
Installation
Follow installation guide here
How to Run
Robot control
The robot control two dynamixel motors via ROS and OpenCR. For providing a easy way of robot control, we built a driving control package name easyGo. See Control Package Page.
SLAM
Mapping should be preceded before deploying robots. See SLAM Page.
Global path planning
Using pcd map, the robot can plan global path using our FMM based modified A*. See GPP Page.
Motion planning
The robot can follow the generated path by motion planner. Our motion planner is using our real-time ground segmentation method named MORP. See Motion Planning (MORP) Page.
Full autonomous driving
RGB-D localization, global path planning and motion planning are integrated in one python script. Just run integration.py. For details, see Integration Page
Human-Computer Interaction
See Human-Computer Interaction Page.
Related Repositories
DPoom_Gazebo
DPoom is also availble in ROS Gazebo simulation with equivalent codes. To simulate DPoom in Gazebo: DPoom_gazebo
Gazebo-CrowdNav
Our navigation method can be simulated in Gazebo. Current state-of-the-art navigation approches based on DRL (CADRL, LSTM-RL, SARL) are available with DPoom platform. To evalute navigation performance with DPoom in Gazebo: Gazebo-CrowdNav
Project Info
2019/12/5 We opened our github repos to public!!.
 |
 |
Purpose of this project
These days mobile robots are rapidly developing in the industry. However, there are still some problems for practical application such as expensive hardware and high power consumption. In this study, we propose a navigation system, which can be operated on a low-end computer with an RGB-D camera, and a mobile robot platform to operate integrated autonomous driving system. The proposed system does not require LiDARS or GPUs. Our raw depth image ground segmentation extracts a traversability map for safe driving of the low-body mobile robots. It is designed to guarantee real-time performance on a low-cost commercial single board computer with integrated SLAM, global path planning, and motion planning. Running sensor data processing and other autonomous driving functions simultaneously, our navigation method performs fast at 18Hz refresh rate for control command, while the others have slower refresh rates. Our method outperforms current state-of-the-art navigation approaches as shown in 3D simulation tests. In addition, we demonstrate the applicability of our mobile robot system by successfully autonomous driving in a residential lobby.
Our team
Taekyung Kim / DGIST Class of 2020 @ktk1501
Seunghyun Lim / DGIST Class of 2020 @SeunghyunLim
Gwanjun Shin / DGIST undergraduate @shinkansan
Geonhee Sim / DGIST undergraduate @jane79
Platform info
SOFTWARE
- Ubuntu 16.04 LTS
- ROS Kinetic
- Python 3.6
- Python 2.7 on Robot Platform
- Realsense SDK 2
- Tensorflow 1.8.0
- OpenCV
HARDWARE
- Platform Computing Unit : Lattepanda Alpha 864
- Intel Realsense Camera D435i
- Turtlebot3 waffle pi
- Outer HW printed by 3D-Printer
Honors
DPoom won the 2019 Samsung Open Source Conference (SOSCON) Robot Competition. Media coverage
Paper info
Our paper had submitted to IROS 2021