#robotics Logs

Mar 30 2018

01:55 AM z64555: I really should order some servos and make me an armatron. :/
01:59 AM esch: I've been playing with some dynamixel servos the past few days... they are fun ^_^
01:59 AM z64555: esch: welcome to the channel. If you stick around long enough, the guys that know what they're doing will show up eventually
01:59 AM esch: ye
01:59 AM esch: I figured I'll just idle until my raspi reboots for whatever reason
01:59 AM z64555: I'm just an EEEN masquerading as a roboticist :D
01:59 AM esch: lol
01:59 AM esch: I'm doing a 2 year for EET
02:01 AM esch: I'm using NI's vision builder AI, it's pretty straightforward if you look into it
02:01 AM esch: I just watched some vids on youtube and read some help files and away I went
02:05 AM z64555: Vision is for vision recognition stuff, isn't it?
02:05 AM esch: controlling the denso via labview requires a proprietary library though, and it always throws a wrench in our stuff
02:05 AM esch: yeah
02:05 AM esch: all I'm doing with it really is pattern matching based on input
02:05 AM esch: we're making an overly complicated vending machine
02:05 AM z64555: ok, whats X, Y, and rZ in relation to the work space?
02:05 AM esch: we have the camera and robot working decent together
02:06 AM z64555: Is the camera statically mounted or on the arm?
02:06 AM esch: so then when I get a coordinate from the calibrated image that coorelates to the robot workspace
02:06 AM esch: nah it's mounted above it on the cell
02:07 AM esch: fixed position
02:07 AM esch: rZ is determined by vision as well, after it pattern matches and finds center mass of said pattern, it makes a coordinate system to throw out orientation
02:08 AM esch: and I just use that value as rZ on the robot
02:08 AM z64555: Ok, the grabber is probably optimized to pick objects from directly above, meaning that if it tried to pick them at an angle it'll have reduced weight limits
02:08 AM esch: true enough but it's fun sized candy vs industrial robot
02:08 AM esch: and the grabber is 3d printed, for whatever reason the school stopped at buying the robot
02:08 AM z64555: *crunch* goes the crunch bars
02:09 AM esch: lmao right
02:09 AM esch: that's why the dynamixel servos, can kinda monitor force via current
02:11 AM z64555: hm, so what is rZ? the yaw of the grabber?
02:11 AM esch: so yeah really what I'm trying to figure out is how to derive rX and rY relative to the X,Y position
02:11 AM esch: yeah
02:12 AM esch: rZ would put the gripper in the proper orientation to pick up the candy bar from above at least
02:12 AM z64555: right
02:13 AM esch: https://photos.app.goo.gl/RKXmE1EYfL1pPOp03
02:13 AM esch: from right before spring break
02:13 AM z64555: Should probably stick with the vertical pickup for most items in the workspace, and only reach out at the far ends
02:13 AM esch: yeah I totally agree
02:14 AM esch: but the problem is it has issues if it gets much closer than where I threw those, or just past the dot grid
02:14 AM esch: so being able to give it a bit more freedom would be great
02:16 AM z64555: ok, you could convert the x-y coordinate into a radius, and check if its above or below some limit, switch to the manual rX (or rY) override
02:18 AM z64555: ...but beyond that I'm not sure what to do. Still trying to learn arm/leg systems myself
02:18 AM esch: yeah I was kind of thinking that, getting an arc of about where that perpendicular reach limit is and define it as a work area, then beyond that give it some sort of angle
02:19 AM esch: but the problem is rX/rY would change as it goes through that arc
02:19 AM esch: so it couldn't be a static value
02:19 AM z64555: of course, you'll have to calculate it
02:19 AM esch: well and I suppose rZ as well
02:19 AM esch: pfft I use labview I can't do maths
02:19 AM z64555: Yeah, you may have to make rZ so that its parallel to the work surface
02:20 AM z64555: I mean use labview to do the maths :P
02:20 AM esch: right
02:20 AM esch: yeah I get what you're saying though
02:21 AM esch: long weekend so I don't get to go back in to bash my head against the robot until monday
02:22 AM esch: I've looking into kinematics since I got home trying to figure out how to derive these values and I just feel like I'm missing something
02:26 AM z64555: at least you're not trying to calculate the forces at the joints needed to achieve all of this :P
02:28 AM esch: ff right
02:42 AM rue_: can I see your model?
04:06 AM Jak_o_Shadows: Is there an explicit solution if you have reundant legs? Like, once your past 3 legs, doesn't getting the force through each leg then turn into a problem with no analytical solution?
04:17 AM rue_: as long as you have a robot to apply it to
04:17 AM rue_: if you have no robot designed, at all, then its kinda hard to resolve an ik soluton for it
04:18 AM veverak: hmm
04:18 AM veverak: OT: well, bullet it is, to try a physics engine
04:18 AM veverak: I hope that I won't feel sorry for that decision
04:26 AM Jak_o_Shadows: I haven't had great luck with bullet
05:37 AM veverak: why was that?
05:40 AM SpeedEvil is now known as Guest71016
05:52 AM newbie72 is now known as micmac
06:24 AM Jak_o_Shadows: I was using it through blender
06:24 AM Jak_o_Shadows: and dunno, the joints weren't as robust as I would hope
06:24 AM Jak_o_Shadows: through, using it through blender isn't ideal probably
06:55 AM veverak: I see
06:55 AM veverak: I want to use directly
06:55 AM veverak: ;)
09:07 AM kd0mto is now known as DagoRed
09:21 AM BitEvil is now known as SpeedEvil
11:22 AM rue_: instead of force feedback actuators, can I use vibrators for force feedback?
11:22 AM rue_: I dont think its the same tho, dosn't give position feedback from the force
11:22 AM veverak: waaait
11:23 AM veverak: how you want to get force feedback from vibrator?
11:23 AM rue_: no
11:23 AM rue_: the game industry seems to have said that instead of pushing back on you, vibration is a fine feedback means
11:24 AM veverak: yeah
11:24 AM rue_: I'm working on force feedback robotics, and need feedback actuators small enough to put in the controls
11:24 AM veverak: yeah, I get it now, hmmm
11:24 AM veverak: interesting
11:24 AM rue_: the mecha is 12' cause thats as small as I can make it
11:25 AM rue_: the control suit has to be the size of a person
11:26 AM rue_: it keeps comming back to me designing a pneumatic valve
11:26 AM veverak: maybe you should ;)
11:27 AM rue_: it might help to work out how much feedback force I need
11:27 AM rue_: cause, as this point I'm trying to forcibly flail the operator
11:28 AM rue_: "the robot fell over and he broke his neck" "was he in it?" "no, force feedback got him"
11:28 AM veverak: :D
11:29 AM SpeedEvil: :)
11:29 AM SpeedEvil: yeah - force feedback is fun, especially if you want the person to stand and move naturally, while supported
11:29 AM SpeedEvil: As it pretty much implies enough force to strain joints if stuff goes wrong
11:30 AM rue_: and I dont need to
11:30 AM rue_: when you hit the limit of the feedback force, you can tell
11:32 AM z64555: it might be from the failures of MS's force feedback system, which had a history of making the joystick go completely limp
11:32 AM rue_: no
11:32 AM rue_: I have a dc motor based feedback demo
11:33 AM rue_: stick on each motor
11:33 AM rue_: you can use either one to control the other
11:33 AM z64555: I mean why the game industry has settled on vibration instead of actual force feedback
11:33 AM rue_: one motor takes force and sends position, the other takes position and sends force
11:33 AM Snert__: cheapicity.
11:34 AM z64555: Vibration is cheap, and also doesn't make the stick into a wet noodle
11:34 AM z64555: or steering wheel, for that matter
11:35 AM rue_: I'm also trying to work out compliant backdriveable linear actuators
11:35 AM rue_: same thing
11:35 AM rue_: I'm wondering about linear brushless motors
11:38 AM SpeedEvil: They are generally much more expensive and heavier.
11:39 AM rue_: but easy to build
11:39 AM SpeedEvil: Very much not.
11:39 AM rue_: liear is
11:39 AM SpeedEvil: Building motors is much more annoying than buying
11:39 AM SpeedEvil: I was wondering about a 'belt winder' type actuator.
11:39 AM rue_: stack some moly magnets and washers for the pin, and stack some steel and wire for the other side
11:39 AM rue_: linear
11:40 AM SpeedEvil: Shit heavy linear.
11:40 AM rue_: but the force feedback would be lumpy
11:40 AM SpeedEvil: Take simple geared brushless motor, feeding to a toothed belt drive cog. Toothed belt pays out and is reeled into an internal spool.
11:41 AM SpeedEvil: As a 'muscle' replacement.
11:41 AM rue_: again, lumpy force
11:41 AM SpeedEvil: It's not lumpy, with the appropriate control.
11:41 AM SpeedEvil: You're not driving it like a stepper.
11:42 AM rue_: https://www.aliexpress.com/item/Neodymium-magnets-ring-50pcs-6x3mm-hole-3mm-n50-rare-earth-industrial-strong-NdFeB-magnets/2019180772.html
11:42 AM rue_: hmmm
11:43 AM rue_: remember this needs to be small
11:43 AM rue_: there are like 30 of the damn things in the hands alone
11:44 AM SpeedEvil: I think you may be overdoing things, unless you want it to be able to tie shoelaces.
11:44 AM SpeedEvil: Which is fine, of course.
11:44 AM veverak: I always doubted that it is actually benefically to make 1:1 control to human
11:45 AM rue_: I do want the 12' mecha to tie its own shoes
11:45 AM veverak: that way it can do only as much as human hand can do
11:45 AM rue_: and I dont mind if it ties my shoes too
11:46 AM SpeedEvil: veverak: 'only' - but the problme with doing more is that you can't do teleoperation well.
11:46 AM SpeedEvil: veverak: For the precise reason you can't teleoperate wings.
11:46 AM SpeedEvil: (and retain arms, never mind neural optimisations in birds)
11:49 AM veverak: yup
11:52 AM SpeedEvil: Well - barring multiple operators, which is of course possible.
11:53 AM SpeedEvil: 'piloted by Dave, and his pet canada goose Fluffy'
12:15 PM rue_: I dont know why I even bother looking at cupons from aliexpress
12:15 PM rue_: there always like "$6 off for every $50000 spent!"
12:20 PM LuminaxWk is now known as Luminax
12:56 PM veverak: :D
01:19 PM mrdata_ is now known as mrdata
02:32 PM ifbizo572: .-. .-.
02:32 PM ifbizo572: .-. .-.
02:32 PM ifbizo572: / \ / \
02:32 PM ifbizo572: / \ / \
03:24 PM esch: rue_: I think you were replying to me earlier, I don't have a model we've just been using the real thing for testing
03:25 PM esch: I think it would be difficult to set up a model for the arm we're controlling since it has to run a program to be controlled via labview
04:08 PM rue_: ok, any data would do
04:08 PM rue_: how many avies?
04:14 PM esch: axis?
04:15 PM esch: it's a 6 axis denso arm, vp 6242
04:15 PM esch: using the digimetrix library for labview control
04:15 PM esch: I can link to all documentation I have on the thing
04:16 PM rue_: so your working on ik for it?
04:21 PM esch: ?
04:22 PM esch: oh
04:22 PM esch: inverse kinematics?
04:22 PM esch: sorry I'm not really familiar with terms, gunna be a bit dense
04:23 PM esch: and yeah that's kind of what I've figured is that I need to figure out my own equation to get a relative angle for the end effector
04:24 PM esch: I just don't know if I have enough points of data for it
04:24 PM esch: for object coordinates I have x, y, and rz of object
04:28 PM esch: I mean really the controller and software handle all of the heavy lifting (thankfully) so I don't have to solve for every axis, but it's easy enough to pull all that data
08:12 PM edmoore_ is now known as edmoore