AWR1642: Questions about the detection accuracy

Amol Vagad

Part Number: AWR1642

Hello everyone,

We have been using the mmwave radar for a few weeks now by mounting it on the front of our development vehicle. Although we have been able to figure out the basic operation of the radar we still have doubts about the detection process as the outputs which we get are many times inconsistent.The details of the configurations are also attached below :

% ***************************************************************

% Frequency:77

% Platform:xWR16xx

% Scene Classifier:best_range

% Azimuth Resolution(deg):15

% Range Resolution(m):0.244

% Maximum unambiguous Range(m):50

% Maximum Radial Velocity(m/s):1

% Radial velocity resolution(m/s):0.13

% Frame Duration(msec):50

% ***************************************************************

sensorStop

flushCfg

dfeDataOutputMode 1

channelCfg 15 3 0

adcCfg 2 1

adcbufCfg 0 0 1 1

profileCfg 0 77 2 5.30 52.68 0 0 15 1 256 6250 0 0 30

chirpCfg 0 0 0 0 0 0 0 1

chirpCfg 1 1 0 0 0 0 0 2

frameCfg 0 1 16 0 50 1 0

guiMonitor 1 0 0 0 0 0

cfarCfg 0 0 8 4 4 0 5120

cfarCfg 1 0 4 2 3 0 5120

peakGrouping 1 1 1 1 255

multiObjBeamForming 1 0.5

sensorStart

I have the following questions depending on a few scenarios :

1. While testing for Adaptive Cruise Control, most of the time we start detecting the vehicle in the front at around 35-40 meters (Max unambiguous range is 50 m). Despite driving in a straight path following the vehicle we see that the detection is sometimes lost in between and gained back again. There have been cases where we have seen the detection being lost even in close distances around 10 m or so. Also, many times the vehicle is not detected at all. What could be reason for such non uniform outputs ?

2. After some reading we discovered that the maximum Doppler velocity is 30 kmph(approx 18 mph). Now we have major problems when we try moving towards a stationary vehicle. Could this maximum Doppler velocity be causing problems that cause us lose detection ? Because we have also experienced that we detect objects quickly and steadily at lower speeds ( around 15 mph) and detection becomes poor at a bit higher speeds ( around 25 mph) in case of running into other stationary vehicle.

3. One of our observations is that we get many detection points when the vehicle moves under a bridge. Is it a drawback for the radar ? If yes then do you suggest some remedies for it ? Also are there any other such factors which affect the radar output ?

4. Sometimes we have also observed that we get many noisy detection points when we move through the areas where we have many objects around ( driving through intersections, across buildings etc ). What can be done to eliminate this ?

5. So far the only method to set configuration parameters on the radar we know is of using the visualizer. We also calculated some values using the estimator and made changes in the parameters like profilecfg, chirpcfg and it helped us improve our outputs. However, is there any other way we could set/ change the parameters ? Because for example we only have maximum range of 50m allowed on the visualizer where as we can use the radar with max range of 80 which could be really helpful for ACC. In your own testing do you have some parameter values for ACC which you would like to share ?

Thanks.

over 8 years ago

0 Vivek Dham over 8 years ago

TI__Guru 58035 points

Hello Amol,
For a 80m ACC kind of application you can improve the performance by some changes in the detection configurations. We will get back to you with more details.

Regards,
Vivek

0 Amol Vagad over 8 years ago in reply to Vivek Dham

Intellectual 445 points

Hi Vivek,

Where can we exactly access the "detection configurations " you mentioned here ?

0 Anil Mani over 8 years ago in reply to Amol Vagad

TI__Expert 4720 points

Hi Vagad,

I was looking at your configuration and there are a couple of changes I'd recommend. The differences are highlighted in the configuration below

sensorStop

flushCfg

dfeDataOutputMode 1

channelCfg 15 3 0

adcCfg 2 1

adcbufCfg 0 0 1 1

% It is better to raise the rx gain to about 40 dB for the ranges (50m) that you are interested in.

profileCfg 0 77 2 5.30 52.68 0 0 15 1 256 6250 0 0 40

chirpCfg 0 0 0 0 0 0 0 1

chirpCfg 1 1 0 0 0 0 0 2

% Increasing the number of chirps per frame helps in improving the SNRs of the target, as well as the velocity resolution, so that the target is more easily detected.

frameCfg 0 1 64 0 50 1 0

guiMonitor 1 0 0 0 0 0

% It is recommmend that you reduce the cfar threshold until the object is consistently detected. However, This can increase the false alarm rate. Play with the threshold until you are satisfied.

cfarCfg 0 0 8 4 4 0 5120

% I have reverted your selection of the cfar parameters to the original set. Is there a specific reason for your selection?. You can change both thresholds by the same amount.

cfarCfg 1 0 8 4 4 0 5120

% Peak grouping is disabled in the range dimension so as to increase the number of detected targets.

peakGrouping 1 0 1 1 224

multiObjBeamForming 0 0.5

sensorStart

multiObjBeamForming 1 0.85

The above changes should improve detection and reliability. There is an app-note called 'Short range radar (SRR) TI Design' which can be accessed from '' which you can also read. This application's configuration is designed for 80m.

Regarding your list of questions.

I have the following questions depending on a few scenarios :

1. Please try again with the above configuration.

2. No, detection sensitivity is theoretically independent of the target's velocity. The 'maximum unambiguous velocity' is simply the 'velocity of the target' at which the velocity measurement by the radar wraps around (i.e, goes from a positive value to a negative value).

3. The best solution (to avoid detecting objects above and below the radar) is to redesign the antenna array so that the elevation field of view (FoV) is minimized. In the default antenna configuration the elevation FoV is approximately 10 degrees.

4. Most detected points are valid (and unlikely to be noise). Aside from redesigning the antenna for the application that you're looking for, you could also implement higher level algorithms to remove unwanted detections by identifying features of the detected points that are un-necessary. For instance, if the detected point is stationary or has low SNR. You could also try and calculate the relative velocity of the object w.r.t to the car and decide whether the reflection corresponds to anything important.

5. Yes. Please checkout the configuration from 'Short range radar (SRR) TI Design'.

Thanks.

Regards,

Anil

0 Amol Vagad over 8 years ago in reply to Anil Mani

Intellectual 445 points

Hi Anil,

Thank you for the detailed response. I will soon try this configuration and get back to you. Meanwhile, I had a doubt regarding the redesigning of antenna array. What do you exactly mean by it and what is the method to carry out the process .

Thanks

0 Anil Mani over 7 years ago in reply to Amol Vagad

TI__Expert 4720 points

Hi Vagad,

Apologies for the late reply.

Different applications require different antennas. For e.g. a long range antenna would have a very narrow field of view and high antenna gain. A short range sensor would have lower antenna gain and larger field of view.

By designing an antenna for your particular set of requirements, you can extract better performance from the radar.

Regards,
Anil

Sensors

Sensors forum

AWR1642: Questions about the detection accuracy