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.

AWR6843: Explanation on MultiObjectBeamForming Algorithm in SDk 3.5 (AOA2DPROC)

Part Number: AWR6843

Hi,

We are working with mmw sdk 3.5 with the 64xx demo code. 

We have a concern with regards to the implementation of multi object beam forming. As per the doxygen present inside the AoA DPU folder, when multi object beamforming is enabled, there must be one or more detected objects per range doppler detections processed. 

In our testing, we have kept a constant moving object at a fixed location and observing that points are more when multi-object beamforming is disabled.

We also tested by injecting same raw data (radar cube) to check the points when multi-object beamforming is enabled and disabled. The result is that both cases are having same points. In some reference input , the enabled one has more points.

Please let us know if there is any more explanation to the algorithm as to why this behavior is observed

Thanks & Regards,

Prathith

  • Hi Prathith,

                     Can you please confirm if you are observing this behavior using the demo code or a TIREX project release?

                      Let me know if you have any questions.

     Thanks and Regards,

    Akshay

  • Hi Akshay,

    I am observing this behaviour on the demo code.

    Regards,
    Prathith

  • Hi,

    The purpose of this feature is to detect multiple peaks for same range/doppler bin detection.

    To simplify, let's assume that there are two identical STATIC corner reflectors. If both CRs are placed in the same range bins, only one will be detected. If we enable the feature, both of them will be detected.

    Thank you

    Cesar

  • Hi Cesar,

    Thanks for the reply.

    We have tested moving objects in static clutter removal enabled mode. We are observing more points when multiObjectBeamForming is disabled. The object is moving constantly at the same location.

    Also, when we keep the multiObjectBeamFOrming threshold = 1. We tried to debug and check if the function is returning single point per range-doppler detection, But to our surprise, it was returning more than one points.

    We are not able to understand as this not an expected phenomena.

    Please let us know more details here.

    Thanks & Regards,

    Prathith

  • Hi,

    Normally, if environment is identical, if ONLY the multiObjectBeamForming is enabled, there should not be more detections.

    You can see in the comments below that there are not new detections when this feature is enabled

    C:\ti\mmwave_sdk_03_05_00_04\packages\ti\datapath\dpc\dpu\aoaproc\aoaproc_common.h

     *    The computed (x,y,z) along with the Doppler value for each detected object are populated in the output
     *    list, structured as @ref DPIF_PointCloudCartesian_t.
     *
     *    If the multi object beam forming feature is enabled, (it can be dynamically controlled from CLI), the algorithm
     *    searches for the second peak in the azimuth FFT output and compares its height relative to the first peak height,
     *    and if detected, it creates new object in the output list with the same Doppler

    Can you share a snapshot of the differences you see?

    Thank you

    Cesar