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?
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?