AWR1843AOPEVM: AWR1843AOPEVM

Hi,

In order to study static obstacle detection I also recommend using the mmWave SDK 3.6 oob demo. The Visualizer of the oob demo includes a Range Profile that helps to understand the environment and noise level.

Ayush Shetty said:

1)How to improve the Accuracy of Cluster coordinates. We were getting a error margin of 10-15cm esp when obstacle is closer to radar(0-40cm)

I assume this error margin is for the range. Is it at 0 deg angle?

Ayush Shetty said:

2)We want to find out the dimension of the Obstacle. Can a cluster size give us the dimension of the Obstacle. If yes, How accurate is it? (please refer attached data below)

No, the cluster size is related to the detected points it does not allow to give the dimensions of the obstacle. With a radar sensor it will be very difficult to determine the size of the object because there is not sufficient angular resolution. There are no sufficient detected points to determine the size of the object.

Ayush Shetty said:

3)How to eliminate the Ghost points when

Type:1.With no obstacles in front of the radar

Type:2 Reflections from the obstacle

First step is to understand where the ghost reflections are originating. Using the range profile in the mmwave SDK oob Visualizer is very useful for this task.

thank you

Cesar

Hey Cesar,

For Question 1.We needed the coordinates of the obstacle but when i place an obstacle at [x,y]->(0,20cm) it shows (0,40cm).

We observe a great error when the obstacle is close to radar.How do we tackle this issue

Question 3.There was constant ghost point observed with no obstacle,we think its due to ground reflection.How can we irradicate this?

Hey Cesar,

For the Carboard box mentioned above,we are attaching the GUI Screenshot.We were expecting the radar to detect multiple points along with edges of box but to the contrary radar is detecting the center of the box.Can you explain exactly why this behavior and could suggest ways to make the radar detect multiple points

(Pls note:Radar height:82.5cm and Box Height is 40cm above ground.)

Warm regards

Ayush Shetty

Ayush Shetty said:

Question 3.There was constant ghost point observed with no obstacle,we think its due to ground reflection.How can we irradicate this?

Hi,

The only way to remove detections from the floor would be to determine the elevation information and filter out the points under a certain threshold.

Ayush Shetty said:

For Question 1.We needed the coordinates of the obstacle but when i place an obstacle at [x,y]->(0,20cm) it shows (0,40cm).

Hi,

If you place the object at 100cm, 80cm, 60cm, 40cm, 20cm, can you please let me know what the detection is?

Ayush Shetty said:

For the Carboard box mentioned above,we are attaching the GUI Screenshot.We were expecting the radar to detect multiple points along with edges of box but to the contrary radar is detecting the center of the box.Can you explain exactly why this behavior and could suggest ways to make the radar detect multiple points

Hi,

Due to the angular resolution of the antenna which is around 30 deg, it will be challenging to detect multiple points. This is the reason why you are only seeing the strongest reflection in the center of the box.

If two static objects are in the same range bin, than they will not be detected separately if the angle between the objects is smaller than the angular resolution.

Thank you

Cesar

We are facing one more issue with obstacle detection:-

A cube(Mild steel)of Dimension 40x40x40cm was kept in front of the radar. For most cases we are getting good detection except for one orientation where the cube is edge-proned i.e The edge of the cube was in front of the radar.I'm attaching a Top-view  pictorial representation of  the curb below:-

1.Could this be happening because the edge is deflecting the chirps or edge has less cross sectional area?

2.Any solution to tackle this issue?

Warm regards

Ayush Shetty

Thank you Cesar.

I'll keep in mind about this limitation. Might use 2 radar at different postion to overcome this limitation 

Warm regards

Ayush Shetty

Hey Cesar,

For a given Obstacle we won't get the Dimension but we wanted to know if we can get point cloud data around the obstacle. For example for a big obstacle we should get lot of point clouds across the whole obstacle but from via radar we are getting a coordinate which is close to the center of the obstacle.

Can we achieve multiple point cloud in obstacle detection, If not is there any lab via which we can achieve this?

Warm regards

Ayush Shetty