• State TI Thinks Resolved
  • Locked Locked
  • Replies 11 replies
  • Answers 2 answers
  • Subscribers 32 subscribers
  • Views 426 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

IWR1843: Detecting objects

Danilo A.
Mastermind 24705 points
Part Number: IWR1843

Hi Team,

We would like to ask your help regarding the customer's inquiry below.

I'm using an IWR1843 radar sensor and I want yo verify the correct working mode of it.

For a test, I run the mmWave demo visualizer with a best range resolution configuration

but when I put a cylindric empty tank of oil, its not detected as a circle, its more like a line

Regards,

Danilo

  • 0
    TI__Mastermind 20160 points

    Hello,

    Please ask the customer to post images of the data collected. This may be the expected response depending on the configuration used. For example, if 2D configuration is selected, then the object will appear as a line. 

  • 0 in reply to Sabeeh Khan1
    Prodigy 90 points

    Hi Sabeeh, i change a little the configuration in the "mmWave Demo Visualizer" tool  from TI and i can find a more suitable point cloud . Despite this, there are some extra points near to the sensor that i cant understand where does it come. For showing what im saying i attached a photo and a video where i identify each of the object and his representation in the occupancy grid and the no sense point in the free space near to the board. Also, i make a yellow rectangle to show the unexpected dots in the free space in front of the sensor. I decide ti share the image like a google photos link for maintain the pic quality.

    In the video you can also see a moving cloud of a person moving in the measurement scene.

    https://photos.app.goo.gl/3jccuH1NmGDLbVpU7

    https://photos.app.goo.gl/SJDVXqqD6JJ3CF3t6 

    Also, looking some TI examples in youtube i find this video where there are not any extra point like my case. ( 1min 02sec https://www.youtube.com/watch?v=8LOyxzPXw8E&list=PLISmVLHAZbTSi8d3wpDH2SJR_FmsGKyz-&index=10 )

    Finally, i would like to know if there are any way to know if my sensor is in good condition because i am starting to thing its damage or something its wrong with it. It has been store in not the best condition and i think it can  has been affected for humidity.

    https://photos.app.goo.gl/JZ3DrHu3sZdTkidT7 

  • 0 in reply to Joaquin Restovich
    Prodigy 90 points

    After lunch i get lunch and answer here, i come back to the measurement field and i see than the problem with the tank its back so i decide to send you a file with the configuration and the results. 

    i expect to use this sensor for an autonomous vehicle and i think this poor poin cloud wont help to avoid obstacles. Im lookingo for a configuration of the sensor that allow me to recognize the shape of the obstacle, so in this case i would like to have a kind of circular cloud in the position of the obstacle. Is it  possible whit this sensor?

  • 0 in reply to Joaquin Restovich
    TI__Mastermind 20160 points

    Hi Joaquin,

    I don't see anything here out of the ordinary. All objects in the field seem to have point clouds associated so you should be able to navigate around them. For a video demonstration of navigation with radar, see our robotics lab: https://dev.ti.com/tirex/explore/node?node=AD79e41r4iMaKIGvWAXKjg__VLyFKFf__LATEST 

    Regarding the tank, yes there are not enough points to determine the circular shape of the tank. But is this necessary for your application? It sounds to me that you would like to detect and avoid. As the sensor moves around the tank, you can start to build a map and see that the tank has a circular shape. 

    For your sensor, oxidation is normal. You can try wiping the antenna with isopropyl alochol and see if that helps, but your antenna does not seem to be damaged. Especially considering the data output. 

    For any points that may be incorrect, you could try reducing the CFAR parameter. If issues continue, then you could try the Rx Gain Phase Compensation method described in the mmWave SDK User Guide. Please ensure that the ground is not the source of these points. 

    In general, I believe your evaluation of the sensor is correct and you are starting to experiment with single-chip radar limitations. 

  • 0 in reply to Sabeeh Khan1
    Prodigy 90 points

    I have seen the video demo using ROS in the pass and its clear that there isn't extra points in the free space and the clouds have a shape similar to the real object shape. For this reason i have doubts about the calibration of my board, I have been changing the CFAR value but in all cases there are points at around 2cm to my board that doesn't exist, its a very height peak of power that i am receiving.

    Making reference again to the video you recommend me, the oil tank is bigger than all the objects, i cant understand whit it appear like a line, there is any explanation you can bring me? 

    Despite all of this, i will read again the  the mmWave SDK User Guide to check the parameters you are talking about

  • 0 in reply to Joaquin Restovich
    TI__Mastermind 20160 points

    Points being near 2cm away from the board does correspond to a board calibration. Please follow the Calibration section after installing the mmWave SDK:

    file:///C:/ti/mmwave_sdk_03_05_00_04/packages/ti/demo/xwr68xx/mmw/docs/doxygen/html/index.html#Calibration_section

  • 0 in reply to Sabeeh Khan1
    Prodigy 90 points

    Despite i follow all the instruction the points near mi board still appear in places where isn't any objects, do you have any suggestions? How reflective the object used for calibration must be? The noisy point has 100dB and my object has 120dB (looking at the range profile) but i don't know if its enough .

    I also want to use this occasion to ask if you are planning to bring a way to use the sensor in ROS 2? Because until now i just find compatibility whit ROS 1  and we would like to use it in a platform with a larger time of support. 

  • 0 in reply to Joaquin Restovich
    TI__Mastermind 20160 points

    Sorry for the delay. If the issues continue, you can try the 'calibDcRangeSig' CLI command. It will average the nearest range bins out. Find more details in Table 1 of the mmWave SDK User Guide. 

  • 0 in reply to Sabeeh Khan1
    Prodigy 90 points

    Hi Sabeeh, thanks for answering.  I have 2 question about the radar:

    1 - I have been changing the value of this parameter and the nearest point disappear. Despite this, playing with this parameter take effect in the rest of the points of the cloud point,  Is it right?

    2 - Do you know if there would be an easy way to integrate the sensor to ROS2 like happens nowadays with ROS1? This is crucial fact because i am starting a really big project and i have to choose a platform for my development.

  • 0 in reply to Joaquin Restovich
    TI__Mastermind 20160 points

    You can edit the parameters to use however much you would like.

    We use the serial package, which does not seem to support ROS 2. http://wiki.ros.org/serial I would suggest you try our solution on ROS 2 and see what issues you run into.