• State TI Thinks Resolved
  • Locked Locked
  • Replies 10 replies
  • Answers 2 answers
  • Subscribers 31 subscribers
  • Views 278 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.

IWR6843AOPEVM: Support Required for IWR6843AOP Testing and Optimization Challenges

Dileep Sharma
Prodigy 130 points
Part Number: IWR6843AOPEVM
Other Parts Discussed in Thread: DCA1000EVM, IWR6843AOP

Tool/software:

Hello TI team, 

We are evaluating the IWR6843AOPEVM for static presence detection and accurate coordinate tracking within a 15' x 15' room. Initial testing has been conducted using the 3D People Tracking binary with the default configuration. However, we have encountered the following issues:

  1. Accuracy and Reliability of Coordinates: The reported coordinates occasionally show inaccuracies, and the reliability of the detected positions needs improvement.

  2. Static Person Detection: The radar often fails to detect a person when they remain static for an extended period.

  3. Ghost Images: The issue of ghost images persists during testing, leading to erroneous detections.

  4. Decreased Data for Static Targets: When a person becomes static, we observe a decrease in the number of detected objects on the UART output, indicating fewer processed data points from the radar.

  5. Maximizing Point Cloud Detection: What specific parameters in the configuration (CFG) file should be adjusted to maximize the number of point cloud data points detected and reported over the UART?

  6. Binary for Raw Data in Spherical Coordinates: Which binary should be used to obtain the maximum raw data (point cloud data) in spherical coordinate format with minimal processing.   
  7. Disabling Grouping by Range and Doppler: How can I disable grouping by range and Doppler for the cloud points using configuration parameters only, to maximize the point cloud data reported on the UART?

We would appreciate your guidance on resolving these issues to improve the accuracy, reliability, and overall performance of the radar for our application.

Regards & Thankyou 

Dileep Sharma

  • 0
    TI__Mastermind 18221 points

    Hi Dileep, 

    I'm looking into your issues/questions. Please give me a day or so to respond.

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 130 points

    Hello Josh,

    No issue I am waiting for your response.

    Regards & Thankyou 

    Dileep 

  • 0 in reply to Josh Dye
    Prodigy 130 points

    Waiting for your answers please asap

  • 0
    Prodigy 130 points

    Hello Ti team, 

    Waiting for your response please suggest the solution ASAP.

    Regards & Thankyou 

    Dileep 

  • 0 in reply to Dileep Sharma
    TI__Mastermind 18221 points

    Hi Dileep, 

    My sincere apologies for the delay, it seems this thread was not appropriately handled while I was out of office. 

    Dileep Sharma said:
    Accuracy and Reliability of Coordinates:

    Can you expand on the extent of the inaccuracies that you are seeing? How often and how much? Can you provide some examples?

    Dileep Sharma said:
    Static Person Detection: The radar often fails to detect a person when they remain static for an extended period.

    You may need to increase the number of frames that a track in the 'static' state remains before being deallocated. This is a parameter which you can modify in the radar configuration. Please refer to this guide, section titled Tracks Disappearing when Static for more information.

     

    Dileep Sharma said:
    Ghost Images: The issue of ghost images persists during testing, leading to erroneous detections.

    Can you explain more the issue you are seeing with ghosts? When does the issue happen? Please also refer to the same guide in the section titled Too Many Tracks (Ghost Tracks) for suggestions of removing ghost tracks. 

    Dileep Sharma said:
    Maximizing Point Cloud Detection: What specific parameters in the configuration (CFG) file should be adjusted to maximize the number of point cloud data points detected and reported over the UART?

    For increasing the number of points that will be detected in the point cloud you can reduce the CFAR threshold for detected points. This will allow points will lower SNR to be added to the point cloud however this may also increase the amount of ghost points/noise that you see. Please refer to section 4.2.3 of the 3D People Tracking detection Layer Tuning Guide for more information on this.

    Dileep Sharma said:
    Binary for Raw Data in Spherical Coordinates: Which binary should be used to obtain the maximum raw data (point cloud data) in spherical coordinate format with minimal processing. 

    The 3D People Tracking binary should be used. This demo already outputs the point cloud data over uart in spherical coordinates. Please refer to the 3D People Tracking user guide for more information on the TLV format. 

    Dileep Sharma said:
    Disabling Grouping by Range and Doppler: How can I disable grouping by range and Doppler for the cloud points using configuration parameters only, to maximize the point cloud data reported on the UART?

     This is not configurable via the configuration parameters.

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 130 points
    Hi josh,
    Please find the more information for your reference below. 

    Accuracy and Reliability of Coordinates:
    We have set up the IWR6843AOPEVM module in a 15'x15' room to measure the coordinates of a person at various positions. However, we observed an approximate error of 2.5 feet in the measurements. These inaccuracies were noted at multiple points within the space.

    Static Person Detection:
    I have reviewed the relevant materials, but the issue persists. Implementing static person detection seems to exacerbate the problem of ghost images.

    Ghost Images:
    When a person remains static for a while and then suddenly moves, ghost images are created. Additionally, the issue becomes more prominent when more than one person is moving in the room.

    Maximizing Point Cloud Detection:
    I adjusted specific parameters but could only achieve a maximum of 150 cloud points on the UART. Could you please clarify the maximum number of point cloud data points that can be captured on the UART per frame?

    Additional Queries:

    1. Is it possible to use more than one radar in the same space, such as in a larger room?
    2. Could you share the experimental data regarding accuracy and reliability that TI has performed recently? I came across some experimental results from TI previously but can no longer locate them. Could you suggest where I might find these?
    3. Is it possible to stream point cloud data directly from live ADC raw data captured with the DCA1000EVM?
    Best Regards,
    Dileep 
  • 0 in reply to Dileep Sharma
    TI__Mastermind 18221 points

    Hi Dileep, 

    Dileep Sharma said:
    Maximizing Point Cloud Detection:
    I adjusted specific parameters but could only achieve a maximum of 150 cloud points on the UART. Could you please clarify the maximum number of point cloud data points that can be captured on the UART per frame?

    The maximum number of points is limited to 150 in the code to improve processing time. You can change this if you wish. This is discussed on this thread: https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/1131357/iwr6843isk-ods-the-max-points-number-limited-to-149-on-the-visualizer 

    Dileep Sharma said:
    Is it possible to use more than one radar in the same space, such as in a larger room?

    Yes, it is possible to use more than one radar and stream the data from both to a separate host. 

    Dileep Sharma said:
    Could you share the experimental data regarding accuracy and reliability that TI has performed recently? I came across some experimental results from TI previously but can no longer locate them. Could you suggest where I might find these

    I'm not sure I understand exactly what data you are looking for. Can you please clarify?

    Dileep Sharma said:
    Is it possible to stream point cloud data directly from live ADC raw data captured with the DCA1000EVM?

    I apologize but I do not understand the question. DCA1000 is used to capture ADC data samples. The point cloud is the result of processing the ADC data and it does not require the DCA1000 for streaming.

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 130 points

    I'm not sure I understand exactly what data you are looking for. Can you please clarify? 

    3D_people_counting_tracing_Data.pdf 

    Other experimental data by TI, similar to what is provided in the above PDF, includes applications such as occupancy detection, tracking, fall detection, static occupancy detection, and people counting using the IWR6843AOP module.

    I apologize but I do not understand the question. DCA1000 is used to capture ADC data samples. The point cloud is the result of processing the ADC data and it does not require the DCA1000 for streaming.

    I know that, but I am just asking if there is any Python script to directly parse the raw ADC data and generate the point cloud data on a Python GUI.

    I have additional questions:

     

    I performed the occupancy test for a person sleeping on a bed, but it was observed that multiple ghost images appeared for a single person, and multiple detected targets were shown on the GUI. The above outputs were obtained after changing the suggested parameters in the configuration file. How can we remove the ghost images?

  • 0 in reply to Dileep Sharma
    TI__Mastermind 18221 points

    Hi Dileep, 

    1. You can find the various experiments and application tests that we have performed in the Radar Toolbox. Check this link for a list of application examples. Check this link for a collection of experiments/tests. 
    2. For parsing and processing of ADC data captured by the DCA1000 there is not python scripts but there is matlab scripts for this purpose. They are provided with the mmWave Studio installation. The scripts are located at:  C:\ti\mmwave_studio_02_01_01_00\mmWaveStudio\MatlabExamples\singlechip_raw_data_reader_example
    3. Which parameters specifically did you tune? I suggest increasing the number of points required to allocate a track and/or increasing the minimum total SNR a set of points must have for track allocation (snrThre and pointsThre of allocationParam)

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 130 points

    Hello Josh,

    I am trying to connect the IWR6843AOPEVM Rev. G with the DCA1000EVM to capture raw data. I have installed all the necessary dependencies and drivers and set up the static IP. However, when I load the BSS FW and MSS FW, the firmware version does not appear at the top of the screen.

    Additionally, when I connect to SPI and click on "RF Power Up," I encounter the following error([RadarApi]:error:read register failed with error -2). Could you suggest possible solutions for this issue?

    Also, could you provide a detailed document on interfacing the IWR6843AOPEVM with the DCA1000EVM? Most of the documents I have found are for different modules. For reference, I am using the latest version of mmWave Studio. 

    Thanks & Regards 

    Dileep Sharma