Data-synchronization-and-collection

This tutorial will teach you how to use ARClab’s desktop-03 to collect data which for neural network training.

Open the JACKAL simulator with camera and city map

[for every rosrun or roslaunch, open a new terminal]

The effect is as follows. The following text will explain how to operate it.

Since training VIO’s neural network requires a car camera and rich scenes, I wrote a camera plug-in based on the simulator that comes with JACKAL. First, open a new terminal, load the camera’s xacro plug-in file:

export JACKAL_URDF_EXTRAS=~/Downloads/zed2i.urdf.xacro

Then open the JACKAL simulator with the city scene (I have embedded the city scene into the simulator scene):

roslaunch jackal_gazebo jackal_world.launch

open a new terminal. Open the camera image to observe the front of the vehicle in real time:

rosrun image_view image_view image:=/right_zed2i/color/image_raw

open a new terminal. Then use the USB cable to directly connect the PS4 controller, and then open the controller control node:

roslaunch jackal_control teleop.launch joy_dev:=/dev/input/js0

Now you can use the controller to control JACKAL and observe the camera image changes:

Keep pressing L1 button and operate the left joystick to move the car (slow mode).

Keep pressing R1 button and operate the left joystick to move the car (fast mode).

Install this ROS function package

Build your own workspace and in your /src folder, run:

git clone --single-branch --branch Data-synchronization-and-collection https://github.com/wisc-arclab/JACKAL_UGV.git

create an include folder in the JACKAL_UGV folder:

mkdir include

then in your workspace root directory (like syn_ws), complie the ROS package by running:

catkin_make

Collect data

After you finished Open the JACKAL simulator with camera and city map part, now we can collect data.

open a new terminal. In syn_ws, source the enironment:

source devel/setup.bash

Open the model_state_stamper node, which will publish ground truth messages with timestamps:

rosrun data_for_deepvio model_state_stamper

open a new terminal. Then create a new_data folder in the ~/syn_ws/src/JACKAL_UGV/data folder (or any folder should be created by yourself anywhere):

cd ~/syn_ws/src/JACKAL_UGV/data
mkdir new_data
cd new_data

run:

rosrun data_for_deepvio sync_node

Then move jackal in the gazebo, and type ctrl+c to finish the collection It will give you synchronized image data, odometry data and imu data (now at 4hz), and will be about 10 imu data between two synchronized data(40 hz).