#robotics Logs

Dec 11 2013

01:00 MrCurious there was an article on how to combine 3 axis accel + 3 axis gyro 3 axis magnetometer into a unified 3 axis combined ensor
01:01 MrCurious giving greater accuract and meaning due to the combined self-corrected system
01:01 MrCurious and i cant remember the keywords to find it
01:22 GargantuaSauce http://www.scribd.com/doc/29754518/A-Efficient-Orientation-Filter-for-IMUs-and-MARG-Sensor-Arrays
14:34 Xenon77 http://picpaste.com/mag013-iOrJbTkK.png
14:34 Xenon77 anyone else got 0.2 V for their output voltage?
15:49 Xenon77 in terms of the formula F = BIL sin theta, what is the value for theta if a force acting on a wire is parallel to the field direction?
15:52 Faek 0
15:53 Faek Wait no. That's when current and B field are same direction
15:55 Xenon77 hmm
15:58 Xenon77 Faek I think your correct
15:59 Xenon77 "The magnitude of the force on the conductor is also dependent on the angle between the conductor and the magnetic field. If the conductor is at right angle to the field the force will be at a maximum while if the conductor is parallel the force will be zero."
15:59 Xenon77 what say you?
15:59 Triffid_Hunter cosine is your friend
15:59 Faek Also F=qv cross B
16:00 Xenon77 Faek, I'm a little confused though, will the value for theta be zero or is it that the force itself will be zero?
16:00 Faek Wel
16:01 Faek Well if the value for theta is 0, that makes sin(0) which equals 0. Hence no force
16:02 Xenon77 ah. ok. thanks
16:03 Xenon77 and if the force acting on the wire is at 75 degrees to the field direction, it will be sin(75) right?
22:23 MrCurious idle hands lead to boredom, boredom leads to ebay
22:26 Jak_o_Shadows So python/numpy arrays are trickier than matlab ones
22:30 MrCurious numpy makes me want to climb back into bed
22:30 MrCurious i would say they are tolerable to use.
22:30 MrCurious likely not super efficient
22:30 Jak_o_Shadows [[[ -0. -0. -100.]
22:30 Jak_o_Shadows [ 0. 0. 97.]
22:30 Jak_o_Shadows [ 0. 0. 0.]]
22:30 Jak_o_Shadows [[ -0. -0. -100.]
22:30 Jak_o_Shadows [ 0. 0. 97.]
22:30 Jak_o_Shadows [ 0. 0. 0.]]]] Is what I have
22:31 Jak_o_Shadows toooo many brackets, and sorry for multi-line
22:31 MrCurious thought numpy had a way of initializing values....
22:32 MrCurious http://docs.scipy.org/doc/numpy/reference/generated/numpy.array.html
22:32 MrCurious i didnt see your np.array()
22:32 MrCurious around that
22:32 Jak_o_Shadows Oh, this is what i'm printing out to check values.
22:34 Jak_o_Shadows Trying to work out the jacobian of an arm. I can get the locations and angles of the arm sections in global coordinates, the difference between the target and those coords.
22:35 rue_house Jak_o_Shadows, dont use matrixies, just do the math you need
22:37 Jak_o_Shadows But bits of it work out so nice..... Also the Jacobian is a matrix?
22:39 Jak_o_Shadows Also, I think i'm doing the jacobian wrong.
22:39 GargantuaSauce if you are only approaching it numerically, yes
22:42 Jak_o_Shadows argh. Brainfade. The vector from points a to b is b-a?
22:42 GargantuaSauce yes
22:42 Jak_o_Shadows good.
22:43 GargantuaSauce also the jacobian isnt reaaalllly a matrix, it's more of a function that is evaluated to produce one
22:45 Jak_o_Shadows It's the rate of change of each arm end with respect to all the joint angles. (presuming only rotational joints in this case)
22:45 GargantuaSauce yes that is what the resultant matrix represents
22:46 GargantuaSauce but the value of that matrix representing the first-order differentials depends on the value of each of those angles
22:47 Jak_o_Shadows Yeah.
22:48 GargantuaSauce k just making sure you got that, cause with that in mind afaics a numerical representation of that matrix isnt super useful
22:48 GargantuaSauce (for viewing)
22:49 Jak_o_Shadows I understood that for IK, one method is just so shove in the current angles, work out jacobian, find the angles it needs to change to, and repeat this until it works
22:49 Jak_o_Shadows I was viewing it to see if I was getting something of the right shape.
23:17 Jak_o_Shadows Hey, it's working :D
23:18 Jak_o_Shadows Not perfectly, but it's not iterating that many times.
23:19 Jak_o_Shadows Bugger that's cool
23:21 GargantuaSauce fwiw i believe that technique is called (or analogous to) gradient descent
23:32 Jak_o_Shadows What's cool is that I would only need to do a tiny bit to move it into 3D
23:34 Jak_o_Shadows I should put joint limits in though.
23:34 GargantuaSauce yes those are a good idea
23:35 GargantuaSauce they are quite effective at limiting its ability to tear itself apart!
23:35 GargantuaSauce (not eliminating it though)
23:36 Jak_o_Shadows I'd have to make the servo brackets first.