#robotics Logs

Feb 10 2015

03:25 deshipu boston dynamics again
04:45 katsmeow-afk youtu.be ?
09:07 rue_more its youtubes short url server
11:06 pythonsnake hi
11:29 deshipu hi pythonsnake
11:29 pythonsnake how are you?
11:36 deshipu that's irrelevant
11:37 SquirrelCZECH deshipu:
11:37 SquirrelCZECH nope!
13:58 tsmeow-afk notes conceptnet5 is as borked as all the previous concept
13:59 katsmeow-afk "dog - antonym - cat" is not true, both are mammals, both have fur, both etc
13:59 katsmeow-afk "cat - CapableOf - catch mouse" really depends on the cat, the mouse, and the conditions
14:00 katsmeow-afk "cat - AtLocation - windowsill" really? have you seen the windowsills in mobile homes?
14:01 katsmeow-afk and i found those gems on one page doing one lookup - http://conceptnet5.media.mit.edu/web/c/en/cat
14:02 katsmeow-afk you cannot tell your robot to see how the cat is doing, and it always goes to the windowsill, where your cat never is!
14:02 deshipu katsmeow-afk: I think your idea of what concept is does not match the ideas of the authors :)
14:02 katsmeow-afk "cat - NotDesires - get wet", except for tigers, who love to play in water
14:04 deshipu katsmeow-afk: look up 'semantic field'
14:04 katsmeow-afk http://conceptnet5.media.mit.edu/web/c/en/semantic_field ?
14:06 deshipu https://en.wikipedia.org/wiki/Semantic_field
14:07 deshipu katsmeow-afk: generally those relations are not hard facts
14:07 katsmeow-afk please explain in simple terms what the authors think when they say flatly that cats do not want to get wet
14:07 deshipu katsmeow-afk: treat them more like fuzzy corelations
14:07 deshipu katsmeow-afk: let me make an example
14:07 deshipu katsmeow-afk: what do you mean by stating that you are afk when you clearly aren't?
14:08 katsmeow-afk that is undesireable, isn't it? it makes everything "possible"..... it's possible to find a swiming pool in the back yard, AND the cat!
14:08 katsmeow-afk my attentions were elsewhere
14:08 deshipu it's undersirable for formulating "hard logic" primitive expert systems
14:09 deshipu but it's quite useful for fuzzy analysis of the general shape of a language
14:09 katsmeow i feel it is better to develop the hard facts, and then ease them for certain conditions
14:09 deshipu the problem is that outside of mathematics we don't really have any hard facts
14:09 katsmeow ah, see, i don't care about the language, i care about the real world
14:10 deshipu katsmeow: which word do you refer to?
14:10 deshipu world*
14:10 deshipu as the "real" one
14:10 katsmeow the one where a dog is not the inverse of a cat
14:11 deshipu we will never have any hard facts about "reality", all we can get is increasing certainty
14:11 katsmeow "dog - antonym - cat" is not true, both are mammals, both have fur, both etc
14:11 deshipu but Bayes theorem says that it will never be 100%
14:12 katsmeow i think it is true that no dog has an exoskeleton, while cats have endoskeletons
14:12 deshipu katsmeow: it's irrelevant how they look, if people, when asked to name something that is "not dog" will still say "cat"
14:12 katsmeow ergo, a cat is not the antonym of a dog
14:12 deshipu katsmeow: it's the most different thing in the same semantic domain, you see
14:12 katsmeow "not dog" = "brick"
14:13 deshipu katsmeow: by the same logic, "white" is not the opposite of "black" -- because both are colors
14:13 deshipu but for most humans, for two things to be "opposite", they have to both belong to the same semantic domain
14:13 katsmeow both are animal, a bird is vastly more differeent from a dog than a cat
14:13 deshipu like the set of furry quadrupeds
14:14 deshipu they have to be similar enough to be grouped together, yet different enough to be on opposite sides of that group
14:14 tsmeow realises the error: the constraining domain, nor the human cultural expectations and expereinces, is not specified in these definit
14:15 deshipu katsmeow: yes, that's because I suspect that one of the goals of colelcting this data is to actually get that information
14:15 deshipu katsmeow: they want to know how people divide words into semantic fields, among other things
14:16 katsmeow heh, so if i define the domain of half of OpenMind, Kinstrey's project, as "InYourDeviantMind", all the porn makes sense
14:17 katsmeow "how the people" is a strategy error, as at least half the people do it wrong
14:18 deshipu see, there is no "wrong" in the humanist sciences :)
14:19 katsmeow i cannotmodel an Ai on a population of beings wherein 5% of them may be psychotic, i don't want a psychotic Ai
14:19 deshipu a humanist only needs pen and paper to do work
14:19 deshipu a mathematician also needs a trash bin
14:20 katsmeow why do they want to know how people divide words into semantic fields?
14:21 deshipu it tells you a lot about a culture
14:22 tsmeow thinks of all the derisive things to say about 80% of the human culture, in gen
14:22 deshipu it's also interesting how it changes in time
14:23 katsmeow yeas, like a Dutch typo gave us "busy", 500 years ago, and we keep perpetuating it
14:24 katsmeow i ,, AARRGG
14:25 katsmeow if i wanted true data, apparently the opposite of conceptnet, where would i go?
14:25 deshipu wolfram alpha?
14:25 deshipu wattson?
14:25 katsmeow those don't give any more data than is in wikipedia
14:26 deshipu maybe because we don't have it
14:26 deshipu there is also wordnet
14:26 katsmeow time for a wolfram beta anyhoo
14:27 katsmeow it must be available in all the online boks
14:27 katsmeow books
14:27 katsmeow boks don't have words
14:29 katsmeow where will i find a cat? ...... Wolfram|Alpha doesn't understand your query
14:29 deshipu http://wordnetweb.princeton.edu/perl/webwn?s=cat
14:31 katsmeow cats can roar, females may roar during coitus, but it's generally considered pedestrian and discouraged
14:31 katsmeow the wordnet version has been in dictionaries for decades
14:33 deshipu it's an old projcet
14:33 deshipu project*
14:33 katsmeow now i know why i see "surfaceText": "You are likely to find [[a rosebush]] in [[a backyard]]" ..... /s/contributor/omcs/
14:34 katsmeow all respects to Kinstry and Sing, but mining morons just leads to moronic data
14:35 katsmeow sadly, i think that was their undoing
14:36 tsmeow looks for someone else's hair to pull
14:36 katsmeow i broke wolfram
14:37 deshipu how could you
14:37 deshipu I was using it :(
14:38 katsmeow "where will i find a cat?" and after 5 minutes it gives me 10 pages about the character "i"
14:40 deshipu wolfram alpha is more about physics, really
14:40 katsmeow even this is much more productive : https://duckduckgo.com/?q=%22cats+are+usually+found%22
14:40 deshipu try http://www.wolframalpha.com/input/?i=how+much+does+a+cat+weight%3F
14:41 deshipu hmm, "no data"
14:41 deshipu but at least it understood the question
14:42 katsmeow it's impressive wolfram uses "i=" (interogative) vs "?=" :-)
14:44 katsmeow "when a tag is missing, it is impossible to know whether it is in fact inappropriate"
15:06 SquirrelCZECH deshipu: https://github.com/SquirrelCZE/Spider
15:06 SquirrelCZECH to get 6 dof, the idea is
15:06 SquirrelCZECH that you gat "start" point -> point where arm is connected to chassis
15:06 SquirrelCZECH "end" -> where you want the foot to be
15:07 SquirrelCZECH and you calc servo positions so it fits in-between
15:08 SquirrelCZECH and, idea how to get actuall 6 dof, is that I calculate new "start/end" combination by moving start point ti [0,0,0] and [0, 0, 0] rotation
15:08 SquirrelCZECH *to
15:08 SquirrelCZECH and also moving end point accordly to it's start point
15:09 SquirrelCZECH so math is there same, like when you count on the fact that start axis is concurrent to "z" axis
16:31 deshipu SquirrelCZECH: I think that individual servo angles are easier done with this http://sheep.art.pl/Inverse%20Kinematics
16:31 SquirrelCZECH yeah
16:31 SquirrelCZECH I used that
16:32 SquirrelCZECH no, I tried to figure out and than I found out I had similar solution
16:32 SquirrelCZECH to angle "first"/"base" servo is easy
16:32 SquirrelCZECH and than it's just triangle :)
16:32 deshipu yeah
16:34 SquirrelCZECH anyway
16:34 SquirrelCZECH in the 3d simulations it shwoed that it works for 6dof :)
16:47 deshipu yeah, now try with 2dof legs :)
16:48 SquirrelCZECH yeah
16:48 katsmeow-afk [OT] http://www.ft.com/cms/s/0/d61b711c-b148-11e4-831b-00144feab7de.html ... Chinese hackers attack blue-chip groups via Forbes website
16:49 tsmeow-afk expects soon: Chinese hackers attack 3D printer, make robot that locks owners out of their h
16:51 deshipu reminds me of that case of perambulating hard disk drives that barricaded themselves in the server room
16:52 katsmeow-afk Walking Winchesters?
16:54 katsmeow-afk people today just don't know of the days when you had to bolt the hdds to the floor
16:56 Tom_itx would a 1n914 work for a emf diode on a 5v relay?
16:56 Tom_itx too small?
16:56 katsmeow-afk might pop
16:56 katsmeow-afk put a cap across it?
16:57 katsmeow-afk something small, 0.1 - 1.0 ufd
16:57 Tom_itx cap instead of the diode?
16:57 katsmeow-afk no UF4001 in stock?
16:58 katsmeow-afk cap and diode
16:58 Tom_itx yeah i have a shitload of those too
16:58 Tom_itx 1n4001 1n4004
16:58 katsmeow-afk the cap slows down the wavefront, take the initial current rush down some
16:58 deshipu also, pops louder ;)
16:58 Tom_itx parallel with the diode?
16:58 katsmeow-afk 1N are lots slower than UF
16:59 Tom_itx well that's the best i got i think
16:59 Tom_itx i'll do some diggin
16:59 katsmeow-afk cost is comparable in qty, i don't buy 1N400x any more
17:00 katsmeow-afk it's like "don't use 1N400x unless it must be slow"
17:01 Tom_itx i'll see what else i have here
17:01 Tom_itx i wonder if i need one across both of them knowing how i'm gonna wire this up...
17:02 Tom_itx you don't think a 4004 will clamp fast enough?
17:03 katsmeow-afk add the cap across it too
17:03 Tom_itx 1uf?
17:03 katsmeow-afk ok
17:04 katsmeow-afk you should run a pulse generator against a 1N400x, and a MDA25xx block, and look with a scope, you'd be surprised at how slowly those diodes come on and turn off
17:05 Tom_itx i believe you
17:05 katsmeow-afk you canhave a inductive kick wipe out a circuit before the protection diodes get a clue
17:06 Tom_itx what about schottky?
17:06 katsmeow-afk good
17:06 Tom_itx 1A 50v
17:06 Tom_itx i have some of those
17:06 katsmeow-afk the relay coil voltage low enough?
17:06 Tom_itx 5v
17:06 katsmeow-afk kool
17:08 katsmeow-afk one way, seemingly seldom used, is to shape the turn-off current/voltage waveform to eliminate the kick
17:27 Tom_itx i don't think it will work
17:27 Tom_itx looks like they centertapped the coil at GND
17:28 Tom_itx i wonder if a cap in parallel with the +5 would make it flip?
17:28 Tom_itx charge the cap enough so when you release power from one side it has enough charge to energise the other side
17:38 Tom_itx nope that didn't work either
18:57 slidercrank brutal treatment of robots. when you're a dog robot and nobody likes you: https://www.youtube.com/watch?v=AZP44vMwYkM stop this violence!
19:13 Loshki Wow, the damn thing is kick-proof and presumably all-terrain. It has military written all over it. It's just two arduinos and an accelerometer, right?
19:16 katsmeow-afk http://youtu.be/M8YjvHYbZ9w
19:16 katsmeow-afk is better
19:25 slidercrank katsmeow-afk, what is that spinning thing? a camera?
19:26 katsmeow-afk i think the description identifies it
19:26 katsmeow-afk my money is on "laser rangefinder"
19:32 Loshki slidercrank: yes, some kind of lidar. Given that google cars are driving around the neighborhood autonomously with similar equipment, it's not surprising.
20:42 rue_shop3 Loshki, I'd like to point out that if there is a revolt by walking robots, its because of being kicked
20:45 Loshki YT video and test scenarios apart, anyone who'll kick a robot 'just for fun' isn't to be trusted with animals, small children, or anything you care about really...
20:47 Triffid_Hunter Loshki: how about someone who'll kick a robot because they're stress-testing some very advanced stability algorithms?
20:47 Loshki Triffid_Hunter: I did say 'test scenarios apart'...
20:48 Triffid_Hunter if someone was selling me a robot designed for running on rough terrain, first thing I'd do is give it a hard time ;)
20:54 codepython7771 Triffid_Hunter: do you know how to handle the map problem on a running robot? obstacles etc?
20:55 Triffid_Hunter codepython7771: SLAM?
20:55 Triffid_Hunter various slam techniques are pretty well described in literature
20:55 codepython7771 Triffid_Hunter: code - that works?
20:56 Triffid_Hunter codepython7771: it's not a drop-in library, you're gonna have to read some math texts.. http://youtu.be/quGhaggn3cQ and http://youtube.com/activevision should give you some ideas of how SLAM can operate
20:57 Triffid_Hunter although they've technically gone full loop into augmented reality
20:57 Triffid_Hunter codepython7771: for SLAM it's incredibly useful to have ground truth to plug into your EKF as well as an IMU
21:00 codepython7771 Triffid_Hunter: should i read that as - it doesnt work ?
21:01 codepython7771 if it doesnt work drag and drop, then it doesnt work yet :)
21:01 Triffid_Hunter codepython7771: you seem to be confusing serious robotics math with arduino libraries
21:02 codepython7771 Triffid_Hunter: Whats the best implementation of SLAM out there that one can use?
21:02 rue_shop3 Tom_itx, does the EMC software let you view the gcode paths?
21:13 Tom_itx yes it has a preview
21:15 anonnumberanon I'm hoping to make a PID in software.
21:15 anonnumberanon Looking through my Controls Systems class notes from a semester long past.
21:15 anonnumberanon far, far away
21:15 Triffid_Hunter anonnumberanon: http://brettbeauregard.com/blog/2011/04/improving-the-beginners-pid-introduction/
21:17 Jak_o_Shadows I looked up conrol systems notes to figure out what the coefficients in Marlin meant for my hot end
21:18 anonnumberanon cool
21:18 anonnumberanon mine is for a quadcopter I'm not sure if i should fuse the gyro and accelerometer as my input to the pid...
21:18 Jak_o_Shadows I needed to figure out what way I needed to change the coefficients to get less overshoot
21:22 Triffid_Hunter anonnumberanon: see what arducopter and paparazzi do
21:25 armyofevilrobots rue_shop3: https://www.dropbox.com/s/rcmuayoxiqmn26g/IMG_20150210_190249.jpg?dl=0 came out within about 0.5mm of flat. There is actually a tiny bit of lift in the middle of the longer stretchers, but I think it’s OK.
21:29 armyofevilrobots 17m:03s
21:31 codepython7771 Great : Now i need a 1538.461 ohm resistor! :)
21:39 armyofevilrobots with a .000006% tolerance!
21:40 codepython7771 :)
23:51 mrdata um, .000006% tolerance? something is very wrong
23:52 Triffid_Hunter mrdata: it was a facetious joke :P
23:53 mrdata i was, however able to get the capacitor tolerance i needed, picking 3 from a collection of assorteds
23:53 mrdata and hooking them up in parallel
23:54 mrdata got 4 digits
23:55 mrdata in the process i found that most of the capacitors in my collection were up to 20% low
23:56 Triffid_Hunter yeah most capacitors are +80/-20% esp electros
23:56 Triffid_Hunter electros are commonly made to be +50% because they drop as they age
23:57 mrdata these were ceramics
23:57 mrdata i guess adding capacitance to tweak a timing circuit is easier than removing it
23:57 Triffid_Hunter ah, the little ones are usually ok, but the higher capacity multilayer ones aren't so great
23:58 mrdata mhm