Jetson Nano SLAM

Background

This project is a Nanyang Technological University (NTU) Undergraduate Research Experience on CAmpus (URECA) under the supervision of Professor Lam Siew Kei.

Refer to the referenced submodules:

• Jetson dual camera
• ORB-SLAM
• Bag tools

These are modules that were forked and modified to fit the project scope.

Aim

This investigation focuses on improving the localization robustness of ORB-SLAM by employing dual cameras and real-time image stitching on the Nvidia Jetson embedded system. The primary objective is to explore the feasibility and effectiveness of this approach in scenarios with challenges such as fast camera movement, limited distinct features, poor lighting conditions, and the presence of dynamic objects.

Specifications

Hardware tested

• Jetson Nano Developer Kit
• Logitech C910 1080P Webcam (2)

Software tested

OS: Linux Ubuntu 18.04.6 LTS
Packages: Nvidia Jetpack 4.5.1-b17, DeepstreamSDK 5.1.0, OpenCV 4.1.0
ROS distribution: melodic

Installation

Clone the full repository

git submodule init
git submodule update

For each of the submodule, ensure that all dependencies are installed and that they function independently. Note that if the Nvidia Cuda tool kit is not installed, you would have to visit here to find a suitable version for your machine.

Usage

Execution

The shell files can be found under the example folder. Ensure that you have execution permissions for each of the file in the folder.

cd Jetson-Nano-SLAM
./example/run

To run each submodule individually, you can run the specific submodule using the shell file, or to navigate to the directory of the submodule.

config

• To change the FPS or the index limit of the image being streamed, navigate to src\srv_tools\bag_tools\launch\img_pub.launch to modify the parameters
• To modify the parameters of the stitched image, navigate to config\dual_usb.yaml