I am using apriltag_ros package to detect the relative pose between my robots. I have a fix tag size (10cm), max detection distance (3.5m) and operating frequency (20 Hz). While keeping them constant, I am trying to reduce the CPU usage since my onboard resources are limited. I am using Jetson Xavier NX as an onboard computer.
So far I have the following observations:
Reducing "native camera resolution" reduces the CPU usage but also reduces the max. distance tags are detected. So I have min. resolution in order to detect the tags at 3.5m.
Reducing "native camera frequency" reduces CPU usage.
Reducing "tag decimation" has similar effect as (1).
I have the following questions:
Are some tags are more easy to detect than others, considering max. detectable distance, e.g. 16h5 ?
If I disable some topics like "Tag Detections", does it reduce the CPU usage?
I need to detect at most 2 tags simultaneously, is there a parameter to limit this in order to reduce CPU usage?
Does "tag_threads" parameter work? Even if I set them to 2, the nodelets always start with max number of threads
Are there other ways to reduce the CPU usage while satisfying my requirements.
Hello,
I am using apriltag_ros package to detect the relative pose between my robots. I have a fix tag size (10cm), max detection distance (3.5m) and operating frequency (20 Hz). While keeping them constant, I am trying to reduce the CPU usage since my onboard resources are limited. I am using Jetson Xavier NX as an onboard computer.
So far I have the following observations:
I have the following questions:
Thank you for your help in advance.
Best, Kagan