Open fwalch opened 9 years ago
This is just an idea I had, not sure if it makes sense. I don't want to complicate things for you or push you to make changes; for the CV module it doesn't make a difference whether we publish to a topic or use tf.
Just FYI, if we use tf, we can use something like the following to publish test transforms (10 Hz in this example):
rosrun tf static_transform_publisher 0 1 1 0 0 0 1 base_link waypoint 10
@noamyogev84 @efiShtain , your call. Try to lower latency as much as possible though.
@noamyogev84 @fselius @efiShtain
After last meeting we discussed how/where the "PID loop" is supposed to run. We talked about that maybe we can just run it in a callback from
/pid_input
, as the PID values need to be updated each time we get new input. One problem with this is that there won't always be continuous input to/pid_input
, e.g. if the CV node loses sight of the target.However, we can use tf to run a continuous loop and keep a fixed rate. So instead of listening to
/pid_input
and receive a callback on each input, the navigation node can poll the last known delta from tf. That way, it can also know if the CV node lost sight of the target, because then the tf data will have an old timestamp.This is how I thought it could work:
What do you think? Could that work?