d2dtracker_jetson

This package can be used to set up the D2DTracker system on Nvidia Jetson paltforms.

Supported Jetpacks

JetPack 5.1.1. Uses NVIDIA L4T 35.3.1
JetPack 5.1.2. Uses NVIDIA L4T 35.4.1
L4T 35.1.0

Quick Setup

NOTE: All th following steps should be done on the Jetson board.

If you face issues with docker permissions after a fresh install of a jetpack, you can execute the following in a terminal on Jetson, and reboot.
```
sudo usermod -aG docker $USER

sudo chown root:docker /var/run/docker.sock

rm -rf ~/.docker

#reboot
```

Clone this repository

cd $HOME
git clone https://github.com/mzahana/d2dtracker_jetson.git

Export GIT_USER and GIT_TOKEN environment variables (contact admin)
```
export GIT_USER=github_user_ID
export GIT_TOKEN=token
```

Setup the d2dtracker container

setup

cd $HOME/d2dtracker_jetson
./setup_d2dtracker.sh

source $HOME/.bashrc
run d2dtracker_container alias to enter the container
[Inside the container] Setup the system

# Enter the container
d2dtracker_container
# Go to the d2dtracker_system pkg
cd $HOME/shared_volume/ros2_ws/src/d2dtracker_system/
./setup.sh

NOTE open_vins is already cloned in the workspace during the setup of the d2dtracker system, and no need to setup a separate docker image for open_vins.

Update d2dtracker docker image

Pull the latest version of this repo

 cd $HOME/d2dtracker_jetson
 git pull origin main

Re-build d2dtracker docker image

 export FROCE_BUILD=true
 ./setup_d2dtracker.sh

Setup nvidia ISAAC ROS packages

fro example, if you wish to test the isaac_ros_visual_slam system, you can build its independent docker image as follows.

Enter the repo's directory and execute the setup_nvidia_container.sh script
```
cd $HOME/d2dtracker_jetson
./setup_nvidia_container.sh
```
NOTE: Provide sudo password when asked.
source $HOME/.bashrc
run nvidia_container alias to enter the container

After entering the container for the 1st time, you need to build the workspace.

cd /workspaces/isaac_ros-dev
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release

Setup the openvins container(optional)

NOTE open_vins can be built inside the d2dtracker-jetson docker image (see below)

Follow these steps if you wish to build an independent docker image to test open_vins package.

setup

cd $HOME/d2dtracker_jetson
./setup_openvins.sh

source $HOME/.bashrc
run openvins_container alias to enter the container
After entering the container for the 1st time, you need to build the workspace.
```
cd /home/vins/shared_volume/ros2_ws
MAKEFLAGS='j1 -l1' colcon build --executor sequential
```
We are using MAKEFLAGS='j1 -l1' as the Jetson may hang!

Run Systems

d2dtracker

Activate the ocntainer alias that was added to ~/.bashrc after you run setup_d2dtracker.sh. This usually done once as it will be loaded automatically when you run a new terminal.
```
source $HOME/.bashrc
d2dtracker_container
```
When you enter the container for the first time after a fresh setup, you will find only once package inside ~/shared_volume/ros2_ws/src. You need to run the setup file to download the remaining packages and build them.
```
cd ~/shared_volume/ros2_ws/src/d2dtracker_system
./setup.sh
```

source the workspace

source ~/shared_volume/ros2_ws/install/setup.bash

Inside the d2dtracker-container, there is a config.sh script inside $HOME/shared_volume. Edit which module you wish to run (set its flag to True), save and close it. Source the config.sh script. Then, run

ros2 launch d2dtracker_system run_system.launch.py

isaac_ros_visual_slam

source $HOME/.bashrc
nvidia_container

To use Realsense camera (e.g. D455), connect it (before you power up the board) and run

ros2 launch isaac_ros_visual_slam vslam_realsense.launch.py

isaac_ros_yolov8

TODO

isaac_ros_dnn_stereo_depth

TODO

open_vins

source $HOME/.bashrc
openvins_container

To test open_vins on ERUOC dataset, you need to download its ros bag, and then play it
in another terminal inside the openvins-container run the openvins estimator

ros2 launch ov_msckf subscribe.launch.py config:=euroc_mav

IMPROTANT: To use open_vins with live IMU/camera system, you will need to calibrate your cameras using Kalibr, calibrate your IMU using allan_ros2 package (only works with ROS foxy, for now), calibrate both cameras and IMU using Kalibr. Then, use the calibration files to update the configuratoin files for open_vins. An example of configuration files for the Realsense D455 camera see the config/open_vins/custom_d455 See below video link for a tutorial.
Reference: https://docs.openvins.com/gs-tutorial.html
Video : https://youtu.be/rBT5O5TEOV4

Automated System Startup

A service is setup to start the d2dtracker system. The service file is called d2dtracker.service and is available in the services directory in this package, here.

This service executes the file /home/hunter/d2dtracker_jetson/services/d2d_service.sh. You need to adjust the path according to your Jetson system setup. However, if you setup your jetson with username hunter and followed the setup instructions in this README, you won't need to change the path to this file.

This script /home/hunter/d2dtracker_jetson/services/d2d_service.sh runs a few commands and then runs another script(/home/hunter/d2dtracker_jetson/services/run_d2dtracker.sh) that starts(or restarts) the d2dtracker-container which itself runs the required launch file inside the container run_system.launch.py located in the d2dtracker_system pkg

For the first time only, you need to run the d2dtracker manually using the alias that was created during the setup process

d2dtracker_container

Once the container is run (and not removed), you can use the service to automatically startup the system.

Enable d2dtracker.service using:

sudo systemctl enable d2dtracker.service

You can disable the service using

sudo systemctl disable d2dtracker.service

Then you can start the service using

sudo systemctl start d2dtracker.service

Then you can stop the service using

sudo systemctl stop d2dtracker.service