IWRL6432AOP: Support Request – Occupancy Detection Coverage Issue with IWRL6432AOP (Motion and Presence Detection Demo)

Part Number: IWRL6432AOP
Other Parts Discussed in Thread: IWRL6432

Tool/software:

Dear TI Support Team,

I am developing an indoor occupancy monitoring device using the IWRL6432AOP radar sensor.

Setup Details:

  • Firmware flashed:
    MMWAVE_L_SDK_05_05_03_00\examples\mmw_demo\motion_and_presence_detection\prebuilt_binaries\xwrL64xx-aop\motion_and_presence_detection_demo_aop.release.appimage

  • Configuration file:
    Based on PresenceDetect.cfg from
    MMWAVE_L_SDK_05_05_03_00\examples\mmw_demo\motion_and_presence_detection\profiles\xwrL64xx-aop\
    (modified as per room size and sensor placement).

Test Environment:

  • Room size:
    Width = 330 cm, Length = 360 cm, Height = 215 cm

    • rangeSelCfg 0.1 4.8
      mpdBoundaryBox 1 -1.8 0 0 1.65 0 2.15
      mpdBoundaryBox 2 -3.6 -1.8 0 1.65 0 2.15
      mpdBoundaryBox 3 -1.8 0 1.65 3.3 0 2.15
      mpdBoundaryBox 4 -3.6 -1.8 1.65 3.3 0 2.15
  • Sensor placement:
    Mounted in wall corner at ~206 cm height, ~10 degree downward tilted and 45 degree for the corner placement.

    • sensorPosition 0 0 2.06 -45 10
  • Complete config file:

% ***************************************************************
% PresenceDetect:  Chirp configuration and Processing chain are 
% optimized to detect any kind of motion, including fine movements 
% (even small movements that are present while sitting still, 
% such as, movement caused by typing, breathing, etc.).  However, 
% there is no velocity measurement reported in this case.  
% It is typically useful for applications such as identifying 
% presence or absence of occupants in an indoor setting.  
% Localization (Angle estimation) of the object is possible.
% ***************************************************************
sensorStop 0
channelCfg 7 3 0
chirpComnCfg 20 0 0 128 1 30 2
chirpTimingCfg 6 28 0 90 59.75
frameCfg 8 0 403 1 250 0
guiMonitor 2 2 0 0 0 1 1 0 0 0 0
sigProcChainCfg 64 8 2 0 4 4 0 .5
cfarCfg 2 4 3 2 0 12.0 0 0.8 0 1 1 1
aoaFovCfg -60 60 -40 40
rangeSelCfg 0.1 4.8
clutterRemoval 1
antGeometryCfg 1 1 1 0 0 1 1 3 1 2 0 3 2.5 2.5
compRangeBiasAndRxChanPhase 0.0 1.00000 0.00000 1.00000 0.00000 1.00000 0.00000 1.00000 0.00000 1.00000 0.00000 1.00000 0.00000
adcDataSource 0 adc_data_0001_CtestAdc6Ant.bin
adcLogging 0
lowPowerCfg 1
factoryCalibCfg 1 0 38 3 0x1ff000
mpdBoundaryBox 1 -1.8 0 0 1.65 0 2.15
mpdBoundaryBox 2 -3.6 -1.8 0 1.65 0 2.15
mpdBoundaryBox 3 -1.8 0 1.65 3.3 0 2.15
mpdBoundaryBox 4 -3.6 -1.8 1.65 3.3 0 2.15
sensorPosition 0 0 2.06 -45 10
minorStateCfg 5 4 40 8 4 30 8 8
clusterCfg 1 0.5 2
baudRate 1250000
sensorStart 0 0 0 0

  • Default UI from the radar tool

Default GUI for the configured zone

UI while detecting occupancy

Test area image: Person sitting at farthest corner (zone 4)

The person sitting at farthest corner

Sensor placement image:

Issue:

  • I am unable to achieve reliable occupancy detection in all 4 zones.

  • Specifically, Zone 4 (diagonally farthest from the sensor) never shows proper occupancy detection.

  • In the current test environment (~3.3 m × 3.6 m room), performance is already limited, and in the actual deployment target (~10 m × 10 m area), I am concerned about reduced efficiency.

Request:

  1. Could you kindly review my configuration and suggest the correct parameter tuning to achieve efficient occupancy detection across all 4 zones?

  2. Please provide guidance on what changes are needed to scale coverage for a larger room (≈10 m × 10 m) with the IWRL6432AOP.

  3. Any best practices for sensor placement (wall vs. ceiling mount, angle, etc.) would also be helpful.

Thank you for your support.

Bhavya

  • Hello.

    We are looking into this; please give us a few days to provide a response.  Thank you for your patience.

    Sincerely,

    Santosh

  • Hello  

    Ok, thank you for the response. Will wait for your update.

    Thanks,

    Bhavya

  • Hi Bhavya,

    There is limitation about placing the box diagonally. You can use the appimage from the below location and attached configuration which will provide appropriate detection. 

    radar_toolbox_3_10_00_05\source\ti\examples\Industrial_and_Personal_Electronics\Motion_and_Presence_Detection\Motion_and_Presence_Detection\prebuilt_binaries\motion_and_presence_3d\motion_and_presence_detection_demo_aop.Release.appimage

    You can use the MPD boxes as designed in the attached configuration

    mpd_boxes_redrawn.cfg

    You can use the same configuration for detection till 10m. You can extend the boxes for 10m usecase

    Regards

  • Hi  ,

    I have tested the configuration file you provided; however, there has been no noticeable improvement in the occupancy (presence) detection. I recorded a video to demonstrate the issue, but due to file size limitations, I am unable to upload it here. Instead, I am sharing a link for your reference:

    Video Link

    Our application requires certain actions to be triggered based on the presence of people within the zone. I would appreciate it if you could review the video and guide me on the next steps.

    Best regards,
    Bhavya

  • Hi Bhavya,

    Thanks for sharing the video. You can configure the thresholds based on the point cloud detected to increase sensitivity.

    You can try with the config attached below

    mpd_boxes_redrawn_v2.cfg

    rangeSNRCompensation 1 5 10 4 5 -20 20 -90 90 3

    You can increase or decrease the the threshold in rangeSNRCompensation (third parameter "10") to increase or decrease the sensitivity.

    Lets us know if you are able to see consistent detection in all the boxes. Do note that the sensor detects only the motion. If the person is absolutely still then the person will not be detected

    Regards

  • Hi  ,

    Thank you for sharing another configuration file. I have tested it and also experimented by increasing and decreasing the third parameter with different values. I observed that detection sensitivity improves when the parameter is decreased, but the overall improvement is limited.

    Our primary use case is to detect presence (occupancy), not just motion. This includes scenarios where a person may be sitting still, sleeping, or otherwise stationary. So please guide on that.

    Moreover, in the current sensor placement scenario, a person positioned at the extreme right or left of the sensor is not being detected. To address this limitation, we have been experimenting with corner placement. However, you had mentioned some limitations regarding this earlier. Could you please elaborate on those?

    Thanks,

    Bhavya

  • Hi Bhavya,

    1. FOV of the sensor is +/-60 degrees. So, any person beyond 60 degrees on either side of the sensor will not be detected

    2.Motion is basically classified as major motion and minor motion. Major motion refers to movement like walking, minor motion refers to small body movements, hand movements, breathing movements. To be able to improve detection of stationary persons, you need to optimize minorStateCfg CLI in the configuration file. You can refer to the below document for details on usage of minorStateCfg

    MMWAVE_L_SDK_05_05_03_00\docs\Motion_Presence_Detection_Demo_Tuning_Guide.pdf

    Regards

  • Hi  

    1.

    FOV of the sensor is +/-60 degrees. So, any person beyond 60 degrees on either side of the sensor will not be detected

    Yes, that's why we were testing the sensor by placing it in the corner of the room to avoid any blind spots. But you mentioned otherwise

    There is limitation about placing the box diagonally.

    So what was that limitation?

    2.

    Do note that the sensor detects only the motion. If the person is absolutely still then the person will not be detected

    Here, you said 'a person who is still will not be detected.' So, will tuning the minorStateCfg help in this case?

    MMWAVE_L_SDK_05_05_03_00\docs\Motion_Presence_Detection_Demo_Tuning_Guide.pdf

    I am referring this guide for tuning but rangeSNRCompensation parameter is not described in it. 

    Thank you,

    Bhavya

  • Hi

    1. There is no restriction on placing the sensor at the corner of the room but software only supports rectangular boxes. If you need diagonal boxes, you will have to modify the software yourself

    2.If a person is making small movements, person will be detected. If the person is absolutely still, then there will not be any detection. Do note that breathing heavily also causes minor motion, so, it is totally scenario dependent. Sensitivity of detection is controlled by minorStateCfg 

    3. For rangeSNRCompensation, you can refer radar_toolbox_3_10_00_05/source/ti/examples/Industrial_and_Personal_Electronics/Video_Doorbell/docs/Video_doorbell_users_guide.html

    Can you also try with the below appimage and configuration files?

    https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/1023/iwrl6432_5F00_aop_5F00_1725.appimage

    mpd_boxes_redrawn_v3.cfg

    mpd_boxes_redrawn_v4.cfg

    Regards

  • Hello  ,

    Sorry for delayed response.

    Thank you for sharing these new config and appimage. Presence detection result is improved in this config. But I have observed that it is also detecting fan rotating motion as motion which gives false detection. What needs to be tuned in the configuration parameters to only detect human? And 

    If you need diagonal boxes, you will have to modify the software yourself

    What are the changes we have to do to support corner placement for the sensor. Is there any reference?

    Thank you,

    Bhavya