Skip to content

Latest commit

 

History

History
74 lines (52 loc) · 2.79 KB

Intel-Joule-ADA-Perception-Demo.md

File metadata and controls

74 lines (52 loc) · 2.79 KB
Raw

Deploy Joule Demos with ADA

Steps to Boot the Joule

  • Make sure the board is securely mounted in the Joule box.
  • Plug in the power jack.

If the system did not successfully boot up, please go through the following steps: - Wait for 50 seconds and reset the MicroSD card > This issue can be solved, if you mount the micro sd card on / and install the system on it.

  • You are then able to SSH to Joule with
ssh jouleu@YOUR.IP.Add.ress

Start Intel Realsense Camera

  • You are able to start the camera by:
cd ~
./run_camera.sh

It has been set default that you need to run roscore on weeboo (IP address is 192.168.2.171) If you are running roscore at other machines with different IP address, please specify the IP address before running the script.

export ROS_MASTER_URI=http://192.168.2.YOUR_MACHINE_IP:11311

If you cannot publish topics, please run:

export ROS_IP="YOUR.MACHINE.IP.ADDRESS"
  • You can modify the camera configurations which defined in the Intel RealSense Camera ROS-Wrapper.
cd catkin_ws/src/realsense/realsense_camera/launch

You may modify the rs_camera.launch file to set the resolution, FPS, enabling camera streams.

Host machine run point cloud demo

source ./devel/setup.bash
roslaunch depth_image_proc cloudify.launch

Then you may use Rviz to view the point cloud data with subscribing to Topic /camera/color/depth/points

Ongoing issues:

  • Color/Depth Synchronization Problem

    The color and aligned depth to color frames may not arrive at the same time. point_cloud_xyz_rgb use message_filter and checking the frame_id to sync two frames. However, it would continuously give the error that two frame ids are not the same, which would not reconstruct the point cloud data.

    We have disabled the commands which checking the frame id. But we are also working on other solutions to make the frame synchronization issue.

    Potential solution would be using custom-defined message type to combine color and depth image together and send together within a message.

  • Using Image Transport

    The local reconstruct point cloud requires Color and aligned_depth_to_color frames. We applied Compressed Image Transport on the color frames and CompressedDepth Image transport on the depth (Aligned depth to color) frames.

    But in order to make the wireless transmission smoothly, we may need to set the dynamic parameters of the transport, like the compression quality.

    #After running the camera on the joule site, open another ssh window to joule.
rosrun dynamic_reconfigure dynparam set /camera/color/image_raw/compressed jpeg_quality 70
rosrun dynamic_reconfigure dynparam set /camera/aligned_depth_to_color/image_raw/compressedDepth png_level 7