#robotics Logs

Mar 15 2018

12:33 AM z64555: hm, I don't think I can actually use euler vectors to define extents
12:49 AM mrdata: what is an euler vector?
12:50 AM z64555: also known as a rotation vector, its a vector that has both the axis of rotation and the angle around said axis
12:50 AM mrdata: https://en.wikipedia.org/wiki/Rotation_formalisms_in_three_dimensions
12:50 AM z64555: yup
12:51 AM mrdata: and what are extents
12:51 AM rue_: and gimbal lock
12:51 AM z64555: extents are hard limits to where a bone/link/etc. cannot go past
12:52 AM z64555: also known as a constraint
12:52 AM z64555: although I may be mixing my jargon up a bit here
12:53 AM mrdata: okay,
12:53 AM mrdata: so in that case to use the extents and euler vectors you'd need some reference direction
12:54 AM mrdata: and orientation
12:55 AM z64555: This is used as part of a kinematic chain, such as an arm or a leg
12:55 AM mrdata: then you'd have to compare a hypothetical rotation for your present circumstnce and test whether it exceeds the limit
12:56 AM mrdata: sure so the reference direction / orientation is found along the chain
12:56 AM z64555: The limits in question are the joint angles, relative to the bone's parent
12:56 AM mrdata: probably at the two bones above and below the joint
01:01 AM z64555: I can define the limits as three angles about the three axis
01:01 AM z64555: but enforcing them... I can't do a simple clamp
01:01 AM z64555: maybe a dot product
01:02 AM rue_bed: is it a model of a real robot?
01:02 AM z64555: we have two models, a biped and a hexpod
01:03 AM rue_bed: so make the model match the mechanics properly and constrain the axies just like the real ones
01:03 AM z64555: well. yeah.
01:04 AM z64555: issue is with determining what math to do when a desired rotation goes beyond the constraints
01:04 AM rue_bed: you stop
01:05 AM rue_bed: and follow the path to the closest point to the target as you can
01:10 AM z64555: so, dot product, clamp, or something else
01:12 AM z64555: maybe a clamp actually would work? The dot product to the axis vectors would effectively squash the input vector to the axis... something I don't want to do
01:53 AM Jak_o_Shadows: Do you have joints with more than 1 DOF?
01:59 AM z64555: yes, most of them are
02:00 AM z64555: except maybe for the hexapod, 2 joints out of the 3 or 4 are 1DOF
02:01 AM z64555: i forget if the contact point/foot is considered a joint
02:51 AM z64556 is now known as z64555
03:59 AM Jak_o_Shadows: physically, how do you have multiple DOF at a single joint?
04:19 AM mrdata: ball joint
04:20 AM mrdata: and at least 3 cables
04:21 AM mrdata: oh that was a question for z64555
04:23 AM z64555: Well, yeah, a ball joint. Even 1 DOF joints have a very minor slip on them, though
04:25 AM z64555: Human forearm/elbox techincally is 1DOF, but can be simulated as having 2
04:25 AM z64555: *elbow
06:30 AM LuminaxHome is now known as Luminax
06:30 AM joga_ is now known as joga
08:52 AM mrdata_ is now known as mrdata
09:16 AM rue_: and how are you actuating the ball joint
09:23 AM z64555: magical pixie dust within the computer
09:23 AM rue_: it'll be easier to simulate if you design it like something real
09:26 AM z64555: I'm not too sure about that, it would involve making the joints into a 3-bar (or is it 4?) linkage with one of them a movable piston
09:27 AM z64555: for a 1DOF joint
11:12 AM squirrel1 is now known as veverak
10:18 PM id4rk_ is now known as id4rk