AWR1843BOOST: The radar when moved towards a standing person, does not detect the person

Hi Abdul,

There is no direct solution for this problem as it depends on a lot of external factors such as the environment and test setup. It is recommended that you play around and fine tune configurations of the gtrack module to achieve necessary results.

Regards,

Kaushik

Thank you for the reply. Could you provide more information on the gtrack module, preferably documentation. How to update its configuration and what are the best possible values for getting the moving radar to detect a standing person?  

Hi Abdul,

There is no guide or information on the gtrack module as you see. You can refer to the CLI commands in the MRR documentation to see what the tunable parameters are. You can refer to the gtrack source provided with the SDK and the feedback from the outputs on the visualizer to understand how you can tune your config for best results.

Regards,

Kaushik

Hi Kaushik,

Appreciate your response. I found in the resource explorer's MRR user guide section that the config is hard coded for MRR module and the file mrr_config_consts.h can be updated to tune the config then the project can be rebuilt to generate the binary. I tried understanding the MRR module and updated some code in the some of the files and now I am getting myself detected as a blue dot(USRR cluster) when I am standing very near like less than 1 meter to the moving machine. Just need to figure out why its not being sent to the tracker so that I get the rectangle around me that will be tracked. Attached images has info on all this.

Also, regarding our background, our company works in railroad domain. So we have just one dedicated path of track, our environment is different from roadways also because their could be people standing on the railway track facing other side of the train and for this case it is critical for the radar to detect that standing person and avoid collision. This radar detects and tracks big immovable objects like other trains but detecting a standing person is also one of our project requirement, I believe the sensitivity setting could be adjusted to make it more sensitive, in any of the ways like I tried in software. My manager and the team really like your product, it performs very well in terms of tracking even for longer ranges. Its better than our current radar provider in many ways. But this one issue is troubling us in finalizing TI as our future radar provider. I like the hardware and software design that TI has built and I know that this product is very much capable of achieving what we are looking for. I will continue my research and in the meantime if you or anyone from your team can assist us in solving this problem then I would be really grateful.

Thanks,

Abdul

Hi Abdul,

Thank you for providing detailed information of your usecase. Although you need to confirm the RF-frequency bandwidth regulation for railway usecase, wheather 77GHz or 60Ghz is allowed. 

TI has both of mmwave sensor variants 77/60GHz.

xWR1843: 77GHz, xWR6843: 60GHz

.

Abdul Shaik said:

e found an issue where if the vehicle is moved towards a standing person then the radar does not detect the person.

Now, when you say person is not detected while radar is moving towards person. what is behaviour when person is moving towards static sensor?

You are not able to detect that person or not able to track it (in MRR application)?

Suggestion: better to use MRR Beamforming application for your usecase, where all 3Tx of AWR1843 are used to create a beam (and steer it), this will give you extended range and sensor can focus towards specific point (where you see any person/object).

https://dev.ti.com/tirex/explore/node?node=A__AJXx96rI7gxMSeZVEpNFrw__radar_toolbox__1AslXXD__LATEST

It is enhanced version of basic Medium range radar application where along with Beamforming feature, HWA is used to do the most of FFT processing to offload DSP.

 Please go over the user guide document to know more on this application.

.

.

Regards,

Jitendra

Hello Jitendra,

Thank you for the reply. FCC regulations allow the usage of 60GHz and 77GHz frequencies, so its not a concern in railways which requires following FCC guidelines.

Now the results observed in three different scenarios for both sdk modules are mentioned below:

1.  Static radar + moving person => Excellent detection and tracking.
2.  Moving radar + moving person => Excellent detection and tracking.
3.  Moving radar + static person => No detection or tracking (although in MRR after I changed some code that I mentioned above, I as a static person was inconsistently detected as a USRR object as a blue dot, but still not tracked).

I tested MRR beamforming and the results are same like we observed in MRR module. I want to update the code like I did in MRR and retest but the beamforming project is not listed in CCS project explorer and the steps in the link that you mentioned, considers MRR module in the instructions for building the beamforming project in CCS.

Regards,

Abdul 

Hi Abdul,

Can you give more info about the scene such as:

• Are there any large metallic surfaces or objects which could mask the signals from the person nearby?
• Speed at which the radar was moving.
• Range of the static person from the sensor
• Angle at which the static person was at
• Have you played around with CFAR thresholds to increase the detection probability of a person?

Regarding the MRR beamforming, you can import the CCS project from the toolbox in Code Composer Studio and follow the same steps.

.

Regards,

Kaushik