-
Notifications
You must be signed in to change notification settings - Fork 1.6k
Description
Hello! Recently when using the collision monitor I discovered that it adds too much latency to the cmd_vel path. The problem to that was waiting for the transform from data source frame to base frame on every callback. I saw that you are using an Advanced API of lookup_transform which relies not only on getting the transform data_source_frame->base_frame but also at fixed_frame->base_frame. This in our case adds latency (we publish fixed_frame->base_frame at 30Hz which adds at most ~33ms to the latency) Publishing fixed_frame->base_frame faster will reduce the latency but i feel like this is more of a workaround rather than a solution.
I wanted to ask whether this is expected and collision monitor could provide such default latency. Maybe you have some thoughts on how to make it run faster?