• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 1122 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR1843BOOST: 3D pointcloud detection in ROS

Marina Gonzalez
Prodigy 10 points
Part Number: IWR1843BOOST
Other Parts Discussed in Thread: IWR1843

Hello, in our company we are interested in the glass detection capabilities of your mmwave radars, as showcased in this video: https://www.youtube.com/watch?v=dTUEZmp49V4 .

Therefore, we recently ordered your IWR1843BOOST Evaluation Board, that outputs a 3D pointcloud, which we aim to integrate together with the LiDAR present in our robot during the mapping and navigation phase, in ROS.

After performing some tests today (see the enclosed videos), we can observe that the radar is very noisy, and the amount of reliable points obtained (shown in dark red) in contrast to those that we can see on your Youtube video is much lower. What we are showing is a top view of the pointcloud, for easier visualization of the distances obtained.

Therefore, could you please provide us with some advise on which parameters to tweak/set in order to increase the number of "reliable targets/points" that we obtain at the output of the pointcloud, to get an output more similar to yours? Or anything that we can do to get something more in the lines of your video would be highly appreciate it.

Some more information about our radar and its configuration:

Model: IWR1843 2.0 Rev C

Flashed SDK: /home/<user>/ti3.5/mmwave_industrial_toolbox_4_10_0/labs/Out_Of_Box_Demo/prebuilt_binaries/xwr1843_mmw_demo.bin

Then, following the instructions on this website https://dev.ti.com/tirex/explore/node?node=AJVkbvjyhr4p7F6L5Elt4w__VLyFKFf__LATEST, we launch the pointcloud as: https://dev.ti.com/tirex/explore/node?node=AJVkbvjyhr4p7F6L5Elt4w__VLyFKFf__LATEST

Thanks in advance. Looking forward to hearing from you soon.

Kind regards,

Marina.

  • 0
    TI__Expert 7510 points

    Hello,

    I took a look at the videos that you shared. From the visualization it appears that you are using the ROS driver. Please confirm. The Youtube video that you linked comparing to LIDAR was done with our Sense and Avoid Demo. Keep in mind the visualization node for the Sense and Avoid demo is different from the one used in the ROS driver. I checked that the configuration files are the same for both demos, so performance should be the same. Could you share a snippet of your .cfg file. For the ROS driver this is located at the following path:

    <REPO_INSTALL_LOCATION>/mmwave_ti_ros/ros_driver/src/ti_mmwave_rospkg/cfg

    Once I get that piece of information and do some confirmation on my side I can make some suggestions for next steps.

    Best regards,

    Connor Desmond

  • 0 in reply to Connor Desmond
    Prodigy 10 points

    Dear Connor,

    Thanks for your quick response and all the information.

    I followed your instructions, by launching the ti_mmwave_sensor.launch from the case of the Sense and Avoid Demo instead of that from the ROS Driver one. The results have now improved, such that our pointcloud is indeed more similar in shape to that showcased in your video demo. The amount of points we get however, is still very far below yours. The accuracy of the detections is also worse, but I imagine this will improve after some proper filtering, as in the demo.

    In the following link, I enclose a video showing the results we are getting by placing some obstacles at different distances in front of the radar. The .cfg you have asked for, as well as the ti_mmwave_sensor.launch we used are also enclosed. We'd be very grateful to know which parameters/files to modify in order to obtain a denser pointcloud and improve the results in general. (https://www.dropbox.com/sh/v1xtgrqlhzvnh9a/AAAFCz7_S7jeWO1Ix118F3w-a?dl=0)

    The radar we purchased, as mentioned earlier in the thread, is the IWR1843BOOST. In the Sense and Avoid Demo files, I see that these 3D detections can also be performed with your 1443, 6843AOP and 6843ISK. Does the model of the radar highly affect the quality of the detections? If so, which one do you recommend?

    Thanks in advance. I look forward to hearing from you soon.

    Best regards,

    Marina.

  • 0 in reply to Marina Gonzalez
    TI__Expert 7510 points

    Hello,

    I would try playing around with the Decay Time property in RVIZ. This property is associated with the point cloud topic that the ROS driver is publishing to. The RVIZ application just subscribes to this topic and it is within RVIZ that you will change this property value. Please refer to the following resource for more information:

    http://wiki.ros.org/rviz/DisplayTypes/PointCloud

    Additionally there is another thing that you can try. The CFAR threshold determines what points actually are considered "detected" thus send over UART. If you lower this threshold then you will get more points, but a false detection(s) become more probable. Alternatively, if you increase it you will get less points, but you might not get a detection for an object that is in your scene. Please refer to the following resource for more information on this command:

    SDK User Guide: Section 3.4 cfarCfg
    <MMWAVE_SDK_INSTALL_PATH>\docs\mmwave_sdk_user_guide.pdf

    Best,

    Connor Desmond