IWR1443BOOST: The Performance for the ROS visualizer lab

Hi Kuerbis,

Assuming that you are using the latest ROS Point Cloud Visualizer Lab from mmWave Industrial Toolbox 2.0.0, I have provided some of the answers below:

1. The ROS Point Cloud Visualizer Lab uses the IWR1443BOOST or IWR1642BOOST EVMs. Based on the antenna simulation plots provided in the respective EVM user guides, the Field of View for the IWR1443BOOST and IWR1642 EVMs is given as 

Azimuth i.e. Horizontal FOV : +/- 60 degrees i.e. total 120 degrees

Elevation i.e. Vertical FOV: +/- 15 degrees i.e. total 30 degrees

2. The sensor front-end configuration (profile_cfg files) used in the ROS Visualizer Lab are provided in the following directory (on the Linux machine where ROS is installed):

~/catkin_ws/src/ti_mmwave_rospkg/cfg

For each configuration, the directory includes a .cfg files and a .PNG file with the same name which shows the Input Parameters used in the mmWave Sensign Estimator GUI to generate the corresponding .cfg file (e.g. Maximum Range, Maximum Velocity,  Range Resolution and Velocity Resolution).

For example, if using the ROS Visualizer Lab with IWR1443 EVM, you'll use the following command to run the visualizer as defined in "Running the Driver" section of the ROS Point Cloud Visualizer Lab user guide.

$ roslaunch ti_mmwave_rospkg rviz_1443_3d.launch

When you run the above command, the system selects the 1443_3d.cgh file from the ~/catkin_ws/src/ti_mmwave_rospkg/cfg directory to configure the mmWave Sensor. This file was generated using the mmWave Sensing Estimator with the settings shown in the corresponding gui file 1443_3d_gui_settings.PNG which is available in the same directory.

I will provide a follow-up response on the Distance and Angle accuracy.

Regards

-Nitin

Hi Kuerbis,

Accuracy (range/velocity/angle) is dependent upon SNR and the corresponding resolution. The actual achieved accuracy would also depend upon implementation factors (FFT size).

The formula for accuracy in general is Accuracy = Resolution / sqrt (2 x SNR). Angular accuracy is dependent upon the algorithm also but for an FFT based algorithm (which is used in OOB demos), this formula can be used.

Considering the IWR1443 configuration used in the ROS visualizer demo (defined in 1443_3d_gui_settings.PNG as exaplined above):

Range accuracy

--------------

Range resolution = 4.4cm

Detection SNR = 9dB

So the theoretical accuracy = 4.4 / sqrt (2x9) = 1.03 cm.

The actual accuracy also depends upon the range FFT bin size which is given by

Max range/FFT size, i.e. for this particular configuration

=> FFT bin size = 9.01 m / 256 = 3.52cm

Therefore, assuming a detection SNR of 9dB, the best case range accuracy for this configuration is 3.52 cm. The accuracy can be improved by using a higher order FFT.

Angular accuracy (Azimuth)

-------------------------------------

Azimuth angular resolution = 15 degrees (i.e. 120 degrees / 8 virtual RX antennas) = 0.26 radians

Detection SNR = 9dB

So the theoretical angle accuracy (azimuth) = 0.261 / sqrt (2x9) =0.06151 radians = 3.52 degrees

The actual accuracy also depends upon the angle FFT bin size which is given by 180/64 (OOB demo uses 64 point FFT for angle calculation) i.e. 2.81 degrees, which is actually better than the theoretical max.

Therefore, assuming a detection SNR of 9dB, the best case azimuth angle accuracy for this configuration is 3.52 degrees.

Regards

-Nitin