• State TI Thinks Resolved
  • +1 person also asked this people also asked this
  • Locked Locked
  • Replies 9 replies
  • Answers 1 answer
  • Subscribers 32 subscribers
  • Views 529 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.

AWR1843: Performance of MRR and beam steering MRR

Francois Charlot
Mastermind 25387 points
Part Number: AWR1843
Other Parts Discussed in Thread: IWR1843BOOST

Hello,

My customer has evaluated the MRR lab from the Automotive Toolbox v3.6. Detection works fine up to 140 meters, but they have an issue with the tracker id consistency which does not seem to be kept during the target's trajectory.

They have then evaluated the beam steering lab from the same toolbox in an almost identical setting. The issue is the detection is much poorer, up to meters only.

Could you please advise on how to solve that? One thought would be to integrate Gtrack into MRR, but before going there, we are wondering whether there is any reason for the beam steering MRR to be less performant than MRR.

The setup differences are the following:

Lab MRR Beam steering MRR
Sub-frames used MRR USRR+MRR
Steering angle n/a -40 to +40°, 20° increment

Thank you.


Best regards,
François.

  • 0
    TI__Guru*** 140784 points

    Hi,

    Beam steering will have better performance only when it is desired to steer the beam at a certain fixed angle. if the steering angle is changed every frame (like in the demo) the performance will not be as good as MRR, especially at angle 0

    Thank you

    Cesar

  • 0 in reply to Cesar
    TI__Mastermind 25387 points

    Hello Cesar,

    Thank you.Customer have set the min and max steering angles so to have the same one across frames. However the outcome is the same as the above configuration.

    Could you please advise? Again, the target is not to use beam forming. It is to get the out-of-the-box detection performance of the MRR lab (which customer are fine with) but with a consistent tracker id.

    Please advise. Thank you.


    Best regards,
    François.

  • 0 in reply to Francois Charlot
    TI__Guru*** 140784 points

    Based on this information the beam forming lab should not be considered since it will not improve the tracking.

    I think the issue here is the detection robustness. Does the customer use/need to multi mode USRR/MRR or only MRR mode?

    Does the customer plan to design their own antenna?

    Thank you

    Cesar

  • 0 in reply to Cesar
    Prodigy 152 points

    Hi Cesar,

    We are currently using only MRR mode. But we are only using that mode due to Trackers are only calculated for TLV type 3. This type is only built when MRR subframe is received, not USRR subframes. Ideally, we would like to have Trackers on USRR subframe as we need to track short distances too.

    Any case, our current performance is good enough, excluding the tracker id consistency. On MRR when a tracker enter in your field view is not followed during its trajectory and could or could not have a unique track ID. That is our current challenge. 

    Regarding the antenna topic, currently we are using IWR1843BOOST EVK and in the future we will use the same antenna as it has the EVK.

    Thank you,

    Javier

  • 0 in reply to Javier Fernandez
    TI__Guru*** 140784 points

    Hi,

    The tracker implemented in the MRR demo is not as robust as the gtrack.

    Please let me check with algorithm team if using the gtrack would solve some of these issues.

    Thank you

    Cesar

  • 0 in reply to Cesar
    TI__Mastermind 25387 points

    Hi Cesar,

    Could you please give us some feedback? Thank you.


    Best regards,
    François.

  • 0 in reply to Francois Charlot
    TI__Guru*** 140784 points

    Hi,

    Sorry for the delay

    If I understand correctly the problem is as follows:

    • an object is moving in the field of view and is assigned a tracker id.
    • the object is not detected for a few frames and disappears
    • the object is detected again and is assigned a new tracker id

    Is this description correct?

    Assuming this is correct, this is related to the configuration and capability of the tracker to preserve the tracker id when the object is not detected for several frames.

    If the object is not detected for some time the tracker will not be able to preserve the id.

    Thank you

    Cesar

  • 0 in reply to Cesar
    Prodigy 152 points

    Hi Cesar,

    Your assumption is correct, that's why Traffic Monitoring lab (among others) provides the State Transition Parameters. By this way it can be configured when a tracker really disappear when it is lost during a certain amount of frames.

    Any way, this is not the topic here. We are currently using Medium Range Radar laboratory and it does not provide a tracker ID, unique or not. There is no tracker ID.

    Once we will have a tracker ID on Medium Range Radar, we can talk about its consistency during the trajectory.

    Is there any way to implement this functionality on the Medium Range Radar? 

    Best regards and thanks for the answer,

    Javier

  • 0 in reply to Javier Fernandez
    TI__Guru*** 140784 points

    Hi,

    Thank you for this additional details.

    Here is the feedback I have received:

    The demo can be modified to add IDs to the tracker structure. The tracker IDs must be  created/maintained/updated by modifying the Kalman filter functions (in the file Extended_Kalman_Filter_xyz.c).

    Thank you

    cesar