• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 29 subscribers
  • Views 329 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.

AWR1843BOOST: gtrack_step not producing any tracked targets

Erik Larsen
Expert 1095 points
Part Number: AWR1843BOOST

Hello,

Objective: I've been working on getting the gtrack library working on the 1843, using lists of fixed/constant targets.

Problem: No matter what config parameters I use, and the targets I input into gtrack_step(), I can never get any associated tracks. Every measurement point is classified as GTRACK_ID_POINT_NOT_ASSOCIATED after I call gtrack_step().

Additional Info

  • SDK: 03.04.00.03
  • The following dependencies have been satisfied
    • gtrack_log
    • gtrack_free
    • gtrack_alloc
  • The gtrack_create() succeeds successfully, providing a handle to the gtrack instance. All memory allocations (in L3) succeed.
  • I have allocated enough memory space for 2D tracking on the DSS as is described in <SDK>\packages\ti\alg\gtrack\test\common\testlimits.h
  • I've tried using a large number of different config parameters, both with and without the advanced parameters
    • I started off using config parameters as found in "Tracking radar targets with multiple reflection points.pdf" which are also similar to the settings found in "<SDK>\packages\ti\alg\gtrack\test\common\gtrackApp.c
    • I have disabled the advanced settings
    • I have enabled the advanced settings, setting boundary boxes very large, thresholds very low, allowed velocities very high, etc...
  • I have put fixed/hard coded measurement points into gtrack_step(), i.e. constant targets with 0 angle, 0 doppler, 100000 snr, and multiple ranges such as 2m,3m,2.1m,2.2m.
  • I am running at the suggested 50ms frame rate.

I am still unable to get any successful tracks out of gtrack_step(). Each input point is always assigned an ID of 255, which means it was not associated with a track. I have confirmed that if I put N targets into a list of measurement points, I get a list of N assigned IDs.

Please advise on any next steps for debugging.

Thanks,

Erik

  • 0
    TI__Guru*** 148514 points

    Hi

    We have to check with algorithm developer and get back to you early next week.

    Thank you

    Cesar

  • 0 in reply to Cesar
    Expert 1095 points

    Thanks Cesar. I look forward to the response.

    -Erik

  • 0 in reply to Erik Larsen
    TI__Guru*** 148514 points

    Hi,

    Here is the feedback I have received from the developer

    Seems like this use case wants to allocate static target. That is a bad idea the first place. We assume that minimum initial velocity was set to 0. This is the thirst parameter in GTRACK_allocationParams structure.

    Depending on the version of the library and configuration parameter, you may be able to see the allocation. Alternatively, you may try setting the Doppler (of hard-coded points) to non-zero.

    Without knowing more details about the application it is hard to provide further recommendations.

    There is very little reason, to track static targets. If I want to see a chair, I would simply cluster points around, declare a target, find a centroid, and associate with a centroid. No need for Kalman. If I want to see people/moving targets, then ignore the static.

    Static detections are very different from the dynamic ones. They are product of different physical processes; their SNRs have very different meanings.

    My advice: ignore the static if you can.

    Thank you

    Cesar

  • 0 in reply to Cesar
    Expert 1095 points

    Thanks Cesar,

    The response was helpful. We will re-evaluate our test scenarios.

    Thanks,

    Erik