Other Parts Discussed in Thread: IWR6843
Dear TI techical support,
we suspect a MAJOR bug in the firmware of the SDK Demo - 68xx AOP for the TI IWR6843AOPEVM.
We setted up the experiment installing the demoboard perfectly vertical on the edge of a table at 1 mt height from the ground. We used the following parameters set:
Desirable resolution: Best Range Resolution
Frequency band: 60-64 GHz
Frame rate: 10 fps
Resolution: 0,044 m
Maximum unambiguous range: 9,06 m
Maximum radial velocity: 2 m/sec
Radial velocity resolution: 0,26 m/sec
Scatter plot selection
In particular, from an experimental comparison between the raw coordinates of the points collected from the TI mmWave radar and the ground truth provided by an OptiTrack vision system based on calibrated cameras, we observed sudden "jumps" in the trajecories estimated by the TI platform, which in our opinion are not due to the sensor performance, but rather to some bug in the following data processing stage or streaming. Consider that the project was compiled with the 3.2.0.6_AOP version of the SDK running the Industrial Toolbox 4.2.0 as well. No software changes were made by us.
We attach some files to let you understand where the problem is. We tracked a person moving in a wide open room with both systems. The plane of motion is labeled in the following as X-Z (X represents the longitudinal direction with respect to the sensor, whereas Z is the trasverse one)
The attached files 1.pdf, 2.pdf and movie1.mp4 are the results obtained with the pedestrian moving sideways from right to left (and vice versa) at about 4 meters from the sensor. Figures 4.pdf, 5.pdf, and movie2.mp4 are the results with the pedestrian moving along a diagonal path from the lower left corner to the upper right corner of the test area. In all pictuers, the solid lines represent the reference ground truth motion, whereas the dots are the points measured from the radar. Please neglect the delay between the results obtained with the two platforms that are due to limited synchronization accuracy.
From the figures (most notably Figs. 1 and 2) it is quite evident that the sign of measured data suddenly changes. A closer comparison between Figs. 1 and 2 suggests the hypothesis that the measurements related to axes X and Z are periodically swapped.
Could you please help us in understanding the reason of such a strange behavior? We insist on the fact that we did not apply any change to the firmware.
Thank you in advance and best regards,
Alessandro
