Short range radar demonstration using TI’s mmWave sensors

[MUSIC PLAYING] Hi, I am Dan Wang, and
I am system engineer for TI’s mm wave sensors. Today, we are shooting a
demo of short range radar for advanced driver assistance
systems using TI’s 77 gigahertz IF CMOS radar. In this demo, the
sensor is configured to use one of the three transmit
channels and all four receiver channels, with a
range resolution as high as 4 centimeters,
and maximum detection range of 40 meters. Two videos will be shown. In the first video,
the raw ADC data is collected and
post-processed on a PC. In this test scene, the parking
lot is about 24 meters in width and about 36 meters in depth. The radar sensor is placed
three feet above the ground, with a field of view of
120 degree in azimuth. Notice we have cars parked
at various distances. Plot 2 shows the signal
amplitude at different ranges. The green line represents
static range profile. Other velocity profiles
are represented by different colors. The peaks correspond
to objects seen by the radar at various range. In this time instance,
nothing is moving. The reflections seen in
the range of 1 to 6 meters are the ground clutter. Detection at 50
meters is a tree. Multiple detection points
around 25 to 30 meters are from the parked
car on the left. Plot 3 is a heat map
representing information of range and velocity, with
the colors representing signal power in dB. Plot 4 shows the detection
and the classroom results. The red circles are the
detected points corresponding to the targets
seen by the radar. Each rectangle is
a cluster of points belonging to the same object. Plot 5 indicates an occupancy
grid was incorporated velocity information. The grid covers plus or minus
20 meters horizontally and 40 meters vertically, with a
cell size of 10 centimeters by 10 centimeters. The color of the occupied cell
indicates the radial velocity. As the car drives into the
field of view of the radar, it can be effectively
detected along a driving path. We can see that. On Plot 4, the physical geometry
predicted by the cluster is very reasonable. On Plot 5, we see a
very rich point cloud representing the moving cars. Radar can also
detect a pedestrian as she enters the scene. One important
observation is that the micro-Doppler
information is also revealed in a range
Doppler heat map, which is critical for
pedestrian recognition. Notice the radar can detect
a person as she leaves a car and walks away. Detecting a small object in
front of another larger one is a very hard problem. However, with the rich spatial
and velocity information shown in Plot 5, we are able
to separate these two objects. This is the second
real time demo video. The test scene is very
similar to the previous one. Now the data is processed
on a TI TDA3 SoC processor in real time. And the detected points
and a cluster results are sent to the PC
for visualization. We can see that both
pedestrian and cars can be detected and clustered
with reasonably high accuracy. Thanks for watching our demo. For more information, please
visit at the link below. [MUSIC PLAYING]


Add a Comment

Your email address will not be published. Required fields are marked *