So, we are interested on doing joint velocity control, but since (it seems) FRI does not provide a velocity interface directly, what we are doing right now is the following closed loop scheme:
(a) Measure current joint position q
(b) Compute joint velocity qdot = f(q)
(c) Integrate it numerically: q_next = q + dt*qdot, in which dt is the inverse of the frequency we are using.
(d) Send the next joint position q_next using Position mode.
Sometimes there is a weird behavior in which, from exactly the same initial configuration and with exactly the same parameters, it does an unexpected movement from a while and then, a couple of seconds later, it performs the expected motion (i.e, the one that we would expect from simulations). Furthermore, this unexpected motion seems to be always the same: it appears randomly once we start the loop but when it appears, it is always the same motion. So, to summarize, it seems that every 4 trials or so it works as expected 3 times and once it does a weird motion (always the same!) for a while and then it goes back to the normal behavior.
We are running a simple resolved rate controller for pose control and the function f(q) is deterministic, so from the same initial configuration we would expect mostly the same motion. It is important to mention that we tried the same scheme using the SmartServo interface and it was working as expected. We wanted to switch from SmartServo to FRI in order to be able to control at higher frequencies.
We are wondering if there is something wrong with the loop (a)-(d) described above. Maybe we need to change some configuration? We tried different stiffness, and the behavior persists.
Thanks.
So, we are interested on doing joint velocity control, but since (it seems) FRI does not provide a velocity interface directly, what we are doing right now is the following closed loop scheme:
(a) Measure current joint position q
(b) Compute joint velocity qdot = f(q)
(c) Integrate it numerically: q_next = q + dt*qdot, in which dt is the inverse of the frequency we are using.
(d) Send the next joint position q_next using Position mode.
Sometimes there is a weird behavior in which, from exactly the same initial configuration and with exactly the same parameters, it does an unexpected movement from a while and then, a couple of seconds later, it performs the expected motion (i.e, the one that we would expect from simulations). Furthermore, this unexpected motion seems to be always the same: it appears randomly once we start the loop but when it appears, it is always the same motion. So, to summarize, it seems that every 4 trials or so it works as expected 3 times and once it does a weird motion (always the same!) for a while and then it goes back to the normal behavior.
We are running a simple resolved rate controller for pose control and the function f(q) is deterministic, so from the same initial configuration we would expect mostly the same motion. It is important to mention that we tried the same scheme using the SmartServo interface and it was working as expected. We wanted to switch from SmartServo to FRI in order to be able to control at higher frequencies.
We are wondering if there is something wrong with the loop (a)-(d) described above. Maybe we need to change some configuration? We tried different stiffness, and the behavior persists.
Thanks.