#robotics Logs

Oct 08 2014

05:24 wolfmanjm This is what I have it seems to be correct no signs switched I don't think https://gist.github.com/wolfmanjm/1382224dda256466defa I'm thinking when it moves in X the femur should move a little to keep the tibia at the same Z?
05:39 deshipu wolfmanjm: hmm
05:39 deshipu wolfmanjm: are you sure it's within the reachable range?
05:39 deshipu wolfmanjm: it will act weird outside of its range
05:40 deshipu wolfmanjm: for instance, I wasted a day with pozitive z
05:40 deshipu positive
05:42 Jak_o_Shadows I'd spent a couple of days trying to visualise it
05:44 deshipu wolfmanjm: by the way, the servos might need reversing, as each leg may be a mirror image of the others
05:44 deshipu wolfmanjm: so that's a separate setting for each servo for me
05:44 deshipu wolfmanjm: that and trim
09:17 rue_house I have really simple kinematic solutions for anyone who will listen
09:17 rue_house that said I should plug them into some of my robots
09:41 SquirrelCZECH ?
09:43 rue_house all the kinematics can be boiled down to two functions that are easy to use
09:46 rue_house er 3
09:46 rue_house dist dist3d(dist p1x, dist p1y, dist p1z, dist p2x, dist p2y, dist p2z){
09:46 rue_house return sqrt(SQR(p1x-p2x) + SQR(p1y-p2y) + SQR(p1z-p2z));
09:46 rue_house }
09:46 rue_house angle elbow(dist C, dist A, dist B){ // returns angle c
09:46 rue_house return rad2deg(
09:46 rue_house acos(
09:46 rue_house ((double)SQR(A)+(double)SQR(B)-(double)SQR(C)) / ((double)2*(double)A*(double)B)
09:46 rue_house )
09:46 rue_house );
09:46 rue_house }
09:46 SquirrelCZECH sounds complicated
09:46 rue_house angle dir(dist x, dist y){
09:47 rue_house return rad2deg(atan2(y, x));
09:47 rue_house }
09:47 uirrelCZECH just got his hands full with 3d mo
09:47 rue_house no, its easy to do kinematics with those functions
09:48 SquirrelCZECH fuuu
09:49 SquirrelCZECH I hate when 3d model ruins my math because it's not made properly
09:50 SquirrelCZECH but so hard is road for learning computer to generate parts for robots ... *evil laught*
09:52 rue_house do em from hand
09:53 rue_house esp if your not making 20 of your roboit
09:59 SquirrelCZECH well
09:59 SquirrelCZECH Assumption is that if I learn computer everything about making parts what I know
09:59 SquirrelCZECH it will by it's nature be better than me
10:00 SquirrelCZECH and faster
10:00 SquirrelCZECH and stronger
10:00 SquirrelCZECH and harder
10:00 SquirrelCZECH :)))
10:00 rue_house it takes a while to learn to make things by hand, but I think you learn important things along thhe way
10:46 SquirrelCZECH rue_house: goal is to make it work with 3d printing
10:47 rue_house ah, your not coding the kinematics then
10:48 SquirrelCZECH ope
10:48 SquirrelCZECH nope
10:48 SquirrelCZECH ;)
10:48 SquirrelCZECH yet...
10:48 SquirrelCZECH rue_house: but part of the design of part
10:49 SquirrelCZECH is to consider kinematics
10:49 SquirrelCZECH I suppose :0
10:49 SquirrelCZECH )
10:49 SquirrelCZECH :)
10:55 rue_house ooo I shoudl do forward kin with this motion capture set up I made
12:04 wolfmanjm deshipu: if I reverse the motors or any one of them then Z would not work properly I am presuming as Z does what I would expect they are all going in the right direction. It may be a range issue but I have tried several different combinations of settings
12:05 rue_house what is it?
12:05 rue_house does it go down but just kinda shakes trying to go up?
12:06 wolfmanjm no Z works fine, moving the leg in X is not working as I would expect
12:06 rue_house aah
12:06 rue_house in the computer model?
12:06 wolfmanjm no on a real leg
12:06 wolfmanjm quaruped 3DoF
12:07 wolfmanjm usign deshipu algorithm for IK
12:07 rue_house ah
12:32 rue_house how amny axies is each leg?
12:54 wolfmanjm 3
13:49 deshipu wolfmanjm: no idea, really
13:50 wolfmanjm deshipu: ok np, just curious when you move a leg in X you expect it to move the tibia out and the femure a littlebit to keep the tip of the leg at the same height?
13:51 wolfmanjm just making sure I understand what is supposed to happen
14:01 wolfmanjm deshipu: ok got it now I need to reverse both tibia and femur, now X does pretty much what I expect... the -ive Z was also confusing me. a bigger Z moves the leg up and more negative moves it down :) had it revsed
14:01 wolfmanjm so now to build three more legs :)
14:02 wolfmanjm got a lot of 3d printing to do ;)
14:02 deshipu wolfmanjm: great
14:02 wolfmanjm thanks for all the help
14:02 deshipu wolfmanjm: be sure to not make them too long
14:02 deshipu wolfmanjm: rememmber that the knee servo has to carry the whole robot
14:02 wolfmanjm yea I have mg90s with metal gears and they are at the limit now
14:03 deshipu hexapod could be easier in this regard
14:03 deshipu otoh, you still have to carry the whole weight on 3 legs
14:03 wolfmanjm yea I see the ebay price for mg90s have gon eup a lot in just a few weeks since I got the 12 I have
14:04 deshipu I use aliexpress
14:05 wolfmanjm I'd like to get a quadruped working though. I can reduce the femur length if I have to bu tit is already pretty short coxa 45mm femur 35mm and tibia 63
14:05 deshipu they have sg90s for $1.35
14:05 wolfmanjm if I make it any shorter it won't get of the ground
14:05 deshipu sounds pretty much what I have
14:05 wolfmanjm uea but as I understand it sg90s gears strip pretty easily
14:05 deshipu haven't stripped any yet
14:06 deshipu I'm on my fifth robot now
14:06 deshipu it gets addicting
14:06 deshipu I'm now thinking about making a really small one
14:07 wolfmanjm also the stls I have for this robot are set for mg90s, sg90 won;t fit. I'd need to redesign the femur and coxa to fit new servos
14:07 deshipu not using servos, but the stepper motors for camera lenses instead
14:07 wolfmanjm ahhh steppers I know a lot about :)
14:07 deshipu something like http://www.ebay.com/itm/5pcs-2-phase-4-wire-stepper-motor-micro-screw-stepper-motor-Slip-Stepper-motor/151159883864
14:07 deshipu wolfmanjm: really? I know nothing
14:09 deshipu aren't mg90s pretty much the same?
14:09 wolfmanjm those look like fun.. I'm tempted to get some too :)
14:09 wolfmanjm no stepper motors and servos are totally different
14:09 deshipu I mean ms90s and sg90s
14:09 deshipu I know the steppers are different
14:10 wolfmanjm you need drivers for stepper motors like drv8825 or http://www.pololu.com/category/156/a4988-stepper-motor-driver-carriers
14:10 deshipu or I could just program one myself
14:10 wolfmanjm oh no I think sg90 are quite a bit taller than the mg902
14:10 deshipu I see
14:10 wolfmanjm steppers need constant current, then you can microstep them too by using PWM on the coils
14:12 wolfmanjm also steppers are quite slow, most cannot exceed 10revs/sec if that
14:12 deshipu that's fine
14:12 deshipu for a walker
14:13 wolfmanjm I guess Icoul dmix and match the sg90 and mg90 use the sg90 for the least torque joints and mg90 for the most torque
14:13 deshipu I was even thinking about building a linear actuator myself, from a screw and a cellphone vibro motor
14:13 Tom_itx unless you use nice drivers
14:13 wolfmanjm Should get some and see
14:13 Tom_itx i get 80ipm from mine on my cnc
14:13 Tom_itx 40,000 steps/inch
14:14 deshipu but then I would need some kind of position sensing
14:14 wolfmanjm Oh on my 3d printers I can get 500mm/sec from the nema17 I use :)
14:14 wolfmanjm but can;t print at that speed or anywhere near it anyway
14:15 wolfmanjm yea my CNC uses 8mm lead screws so max speed is around 30mm/sec at 1/32 micro steps and 800steps/mm
14:16 wolfmanjm deshipu: yea I played with the magnetic linear position sensors, but the mag strips are hard to get especially if you want accursate ones
14:16 wolfmanjm I did make a DRO for my lathe with one though
14:17 deshipu wolfmanjm: I was thinking that I could maybe measure the resistance of the screw itself
14:17 deshipu wolfmanjm: you know, put one wire at one end, and another at the nut
14:17 wolfmanjm that is an interesting idea :)
14:17 deshipu but it would be probably hopelessly inaccurate
14:18 wolfmanjm closed loop servos systems are kind of fun too but yea you need to be able to measure the position fairly accurately
14:19 wolfmanjm the linear magnetic ones I used have a 12micron resolution which is pretty good
14:20 wolfmanjm they use quadrature encoding
14:20 deshipu another option would be to just use the pots for joints
14:20 wolfmanjm yea I have seen people do that
14:21 wolfmanjm with an inverted pedulum type of setup
14:21 wolfmanjm pendulum
14:21 deshipu could work if I glued them to popsicle sticks
14:21 deshipu you know, the small pots that you set with a screwdriver
14:21 wolfmanjm I never really got my self balancing robot to stabilize though so I gave up on PID control and stuff ;)
14:22 ring0 i'm looking for major robotics companies, aside from kuka. do you have any quick ideas?
14:22 deshipu wolfmanjm: I'm trying to get into that a little now with that MIT course on underactuated robots
14:22 deshipu ring0: boston dynamics? :P
14:23 deshipu wolfmanjm: I will see how much of it I can understand
14:23 ring0 sure an interesting company deshipu
14:24 wolfmanjm I almost got it tho balance but then it started to oscillate and the fast reversal of the motors blew out all my electronics :( they were 12v 2A motors, andf the back emf was probably well over 40v
14:24 wolfmanjm should have used optical isolators or something
14:25 deshipu wolfmanjm: diodes?
14:25 wolfmanjm that was the second MCU I blew so I gave up on that itr was getting too expensive
14:25 deshipu I have a bag of arduino pro minis to burn :)
14:25 deshipu $2 each
14:26 wolfmanjm the motor control chips were supposed to have them built in and the motor power was separate so I have no idea how the back EMF got through the motor drivers into the logic
14:26 wolfmanjm yea I have a couple, but I was using teensy3 and teensy2 which are 10x more expensive
14:27 deshipu ouch
14:27 wolfmanjm teensy3 was needed to habdle the extra computation
14:27 wolfmanjm but they are very sensitive to noise
14:28 wolfmanjm I should get some optical isolators and use them to separate the MCU and logic from the motor controllers
14:30 deshipu I connect the arduino through the serial to the computer and use python to do the computation
14:30 deshipu and recently through a serial bluetooth module
14:49 delinquentme Ok so Im looking to pick up a medical thermister . These are probably going to be analog ... and I need to sort out what kind of circuit I need to build to get it working
14:49 delinquentme What are the things I need to consider? http://www.exacon.com/Global/Global13/Global13FR2.html
15:08 RifRaf wolfmanjm, were your linear encoder modules of the as5311 variety from AMS?
15:09 wolfmanjm RifRaf: yes there is a circuit on thingiverse for them
15:09 RifRaf also those baby steppers you were looking at, have a bunch here, cute but not very much torque if you want to move something like a limb
15:10 RifRaf wolfmanjm, yep i have some pcbs i just got in from OSH park, searching for my soldering station so i can start experimenting
15:11 wolfmanjm did you get the magnetic strips from AMS too?
15:11 wolfmanjm you can;t use any old mag strip
15:12 RifRaf yep have 1 in storage and one new one sitting here for them
15:13 deshipu ring0: even a 2cm limb?
15:14 RifRaf 300mm long, gonna add then to a printer to make more accurate threadless ballscrew system
15:14 deshipu RifRaf: ^^
15:15 RifRaf deshipu, yep if there is any weight above the limb, i used them one a micro reprap, was only trying to turn a 3mm shaft with micro threadless ballscrew
15:15 RifRaf have since moved to steppers about 15mm wide that are doing the job
15:16 RifRaf the 6mm ones are very cute though, will find a use for them somewhere one day
15:16 deshipu RifRaf: I want to make a 2 or 3dof quadruped robot the smallest that I can
15:17 deshipu not much weight, just a 1s lipo and a small pcb with the mcu
15:17 RifRaf deshipu, at the price they cost you should still get some and experiment
15:17 deshipu RifRaf: do you use a special driver for them?
15:17 RifRaf i had to modify the stepper drivers to output very small currents though, easy to kill the motors
15:17 deshipu RifRaf: or do you drive them with software?
15:17 RifRaf i used standard reprap hardware and drivers
15:18 deshipu I see
15:18 deshipu ideally I would like to use software
15:18 deshipu so that I have less stuff in there
15:18 RifRaf modified to output about 20-30mA though
15:19 deshipu hmm, arduino can output 40mA per pin :)
15:21 RifRaf http://i.imgur.com/9Ujoq7N.jpg
15:21 deshipu cute
15:21 RifRaf work time :/ cya
16:40 Hyratel1 ergh, having a logistics-diplomacy heartburn-inducing situation in FRC
21:32 rue_bed-aok dark at 7!
21:42 Tom_itx warmed up the toaster oven tonight
21:43 rue_shop2 ah yes
21:44 Tom_itx fired up the cnc too
21:58 rue_shop2 programmers or holiday presents
21:59 Tom_itx the former
21:59 Tom_itx blue ones