#robotics Logs

Dec 21 2015

03:37 deshipu Snert: you can make a linear actuator from a servo and a spent glue stick :P
03:39 veverak fuuu
03:40 veverak owner of the package I can't compile in ROS is from USA
03:40 veverak daaaamn
03:40 veverak time difference will make things going for a long time
03:40 veverak he is willing to help, so it's still good anyway :)
03:48 deshipu I'm still trying to get servos working on wipy, but it's pretty much hopeless
03:49 deshipu the timers are set up so that you can't get frequecy lower than 611Hz
03:54 veverak so still use arduino as pwm generator?
03:54 veverak :)
03:56 veverak https://www.adafruit.com/product/815 sweet
03:56 deshipu but the wipy has so many pins!
03:56 deshipu veverak: that is for LEDs
03:56 deshipu veverak: it's high-power
03:57 veverak I see
03:58 deshipu you can use it for servos, but it's a waste
03:58 deshipu that's why it's so expensive
03:58 deshipu a pro mini with i2c slave and sevo code is much cheaper
03:58 veverak yeah
03:59 deshipu veverak: https://bitbucket.org/thesheep/tote/src/tip/i2cslave/
03:59 deshipu veverak: a very basic sketch
03:59 deshipu could be greatly improved
03:59 deshipu but good enough for now
04:00 veverak nice union trick :)
04:09 veverak awesome!
04:11 deshipu veverak: cool, can you fix the indentation too, please?
04:11 deshipu have the second line begin at the level of (
04:12 deshipu and spaces around *
04:12 veverak deshipu: like this?
04:13 deshipu much better, thanks!
09:03 anonnumberanon hey deshipu
09:03 anonnumberanon Your servo driver you want to make it accessible serial, SPI, I2C?
09:08 SpeedEvil wifi!
09:08 SpeedEvil ESP8266
09:08 SpeedEvil What could possibly go wrong?
09:10 anonnumberanon nrf24l01 rather, that's the next step over bluetooth, for 80 cents
09:10 anonnumberanon the "not quite wifi" chip
09:11 anonnumberanon 100m range
09:11 anonnumberanon I just have trouble with it right nowl
09:11 anonnumberanon to make it stop working when i want to
09:12 anonnumberanon seems to be doing stuff in the middle of other important timing-sensitive routines
09:13 anonnumberanon My quick hack is to use it on an additional micro and connect that micro to the main micro with serial because I know serial won't interfere with the main micro's program, but I'd rather adapt the nrf24l01 library or calling methods in a way that it would work only when I want it to.
11:20 deshipu anonnumberanon: I will either use i2c or a single-wire protocol similar to what the addresable leds use
11:20 anonnumberanon addressable LEDs thing needs what frequency?
11:21 deshipu it can be as slow as you want
11:25 SpeedEvil err
11:25 SpeedEvil no, many of the addressible LEDs have fixed clock speeds
11:26 SpeedEvil (or nearly so)
11:26 SpeedEvil Oh
11:26 SpeedEvil similar to
11:26 SpeedEvil never mind
11:30 deshipu well, they have fixed length for the zero and one signals
11:30 deshipu but you can send them as slow as you want
12:07 SpeedEvil Well, yes, but 'fixed length for zero and one' means you have to do lumps of data at the average 0/1 speed
12:08 deshipu they have pretty large tolerance, actually
12:08 deshipu anyways, I will start with i2c for sure
12:09 deshipu because that fits my requirements, apart from requiring an additional wire
12:25 SpeedEvil one-wire is useful too
12:25 SpeedEvil I wish they'd sell IDs
12:26 deshipu SpeedEvil: I can always rename it to "single-pin" and sell my own ids ;)
12:35 deshipu the interviews with Sophie Wilson on youtube are pretty interesting
12:52 anonnumberanon I will most likely try and build this after the robot project: https://www.youtube.com/watch?v=KPtsLPMTVnY
12:55 anonnumberanon from scratch though cause those things are thousands of dollars
12:55 anonnumberanon am i cray?
12:55 deshipu you are shrimp
12:56 deshipu didn't the Furze guy build those from toilet utensils?
12:58 anonnumberanon a bunch of people make them in their shops from nothing
12:59 anonnumberanon seems like a cool mechanical engineering project
13:01 deshipu I'm not sure
13:01 deshipu I mean, you can't even bring it to school to show off ;)
13:10 anonnumberanon oh im not at school
13:11 anonnumberanon I would only use it to terrorize the neighbours' pets and other good uses like that.
13:11 anonnumberanon Put two of them on the back of a small Gundam.
13:13 deshipu better use ducted fans with a brushless motor
13:13 deshipu lighter, more powerful, safer
13:15 SpeedEvil I want a nice 3d printer which can do inconel
13:16 anonnumberanon :)
13:17 anonnumberanon go work for SpaceX then, they do this type of things
13:18 anonnumberanon Can't wait for the launchu.
13:22 deshipu meh, I wouldn't even call that an orbit
13:22 deshipu it's all within the atmosphere
13:22 deshipu it really went all downhill from the moon days
13:23 SpeedEvil The second stage is going into orbit.
13:23 SpeedEvil the first is not.
13:23 SpeedEvil In principle, the first could get to orbit - if it had limited payload and no second stage, but that is fairly pointless.
13:24 deshipu ok, it's orbit, but it's not space
13:24 SpeedEvil ...
13:24 deshipu there is still thin athmosphere at the height where ISS is
13:24 SpeedEvil There is atmosphere everywhere.
13:25 SpeedEvil - solar wind, intergalactic, ...
13:26 deshipu right, but this is Earth atmosphere
13:27 anonnumberanon deshipu, dude it's like kinda the golden age of space right now
13:27 anonnumberanon yeah we don't land on moons and planets anymore but the tech we have is nice, and it may happen very soon
13:28 anonnumberanon for the first time NASA this year will finally get all the budget they wanted, with 1 billion + extra dollars.
13:29 deshipu anonnumberanon: compare that to the moon landing budgets
13:29 SpeedEvil Budget comparison is idiotic.
13:29 deshipu it
13:29 SpeedEvil We first need to start reducing launch cost.
13:29 anonnumberanon yeah but the president got shot and all, let's forget about that time
13:29 deshipu it's not me who brought it up
13:29 SpeedEvil So we can reduce cost of hardware because we don't care so much about having to relaunch
13:29 SpeedEvil And so we can relax margins
13:29 deshipu anonnumberanon: it all ended well, I think you really should shoot more of them
13:30 anonnumberanon im french i won't take that responsibility
13:30 anonnumberanon :)
13:31 deshipu I'm a big fan of democration applied directly ;)
13:31 SpeedEvil Anyway - if you don't think rocket-robots landing vertically is cool, ...
13:31 SpeedEvil ~6 hours till launch
13:32 deshipu it's nothing new, really, just the same thing cheaper
13:36 anonnumberanon >landing rockets being nothing new...
13:36 anonnumberanon Dude, sci-fi movies are becoming real. Instead of seeing badass spaceship landings in movies you'll get to see it for real.
13:37 onnumberanon waits 6 hours while learning some
13:38 deshipu anonnumberanon: can't wait for that mad max
13:39 anonnumberanon have you seen the blue origin landing?
13:43 deshipu the one that tipped over?
13:44 deshipu they should really learn to tune their PID ;)
13:49 anonnumberanon lmao
13:49 anonnumberanon no, not that one
13:50 anonnumberanon https://www.youtube.com/watch?v=9pillaOxGCo
13:50 anonnumberanon this one did it from like near space but not space
13:50 anonnumberanon but at least they demonstrated descent and landing success
13:52 deshipu anonnumberanon: that's not even going to the orbit
13:53 deshipu anonnumberanon: we have stratospheric baloons that go higher
13:53 deshipu anonnumberanon: so there is a difference between going really high and then falling right back
13:53 deshipu anonnumberanon: and going to orbit
13:54 deshipu anonnumberanon: and that difference ~8km/s
13:54 robotustra 100 km - not bad
13:55 anonnumberanon it's halfway, enough to start seeing the curvature of the planet or something
13:55 robotustra I would not fly anyway
13:55 anonnumberanon it's 1/10 the difficulty of what may happen tonight
13:56 robotustra do they run something tonight?
13:57 deshipu it's 1/1000th difficulty of actually getting to orbit
13:58 robotustra what are you talking all about?
13:58 robotustra go make your robots
13:58 anonnumberanon the rocket landing that may land back from launching satellites tonight
13:58 robotustra :)
13:58 anonnumberanon just look it up
14:00 shipu goes make his ro
14:44 Hyratel deshipu, at this point "orbit is easy"
14:44 Hyratel all the logic and control has been figured out for 50 years.
14:44 Hyratel apollo made orbit completely automatic
14:47 Hyratel anonnumberanon, ooooh i hadn't seen that vid yet
14:47 anonnumberanon well yeah it just happened
14:48 Hyratel to make orbit all you need is a rocket with enough dV and a half-decent attitude control system
14:48 Hyratel to land is another matter
14:48 Hyratel complicate it by having a Designated Target LZ, an aerodynamically awkward booster, and limited deep throttle capability
14:49 Hyratel I notice the BO Booster actually braked to a half vertically about its own height above the ground and 'waggled' around a bit before setting down
14:49 Hyratel s/half/halt
14:49 Hyratel it had a lot of lateral motion and pendular action going on
14:50 deshipu Hyratel: except you need tons of fuel
14:50 SpeedEvil deshipu: not quite
14:50 SpeedEvil deshipu: the actual landing takes only a couple of tons of fuel
14:50 deshipu SpeedEvil: I mean to make orbit
14:50 SpeedEvil ah
14:50 SpeedEvil yes
14:50 SpeedEvil misread
14:50 deshipu they don't get to the orbit, they just raise very high and fall down
14:50 Hyratel deshipu, that's only logistics. the math and controls have been figured otu
14:50 Hyratel deshipu, I don't think you space much, do ytou?
14:51 Hyratel do you play KSP?
14:51 deshipu Hyratel: logistics is the hard part of space programs :)
14:51 Hyratel Space is hard. Orbit is hard. but it's a known quantity, that is the same hard every time
14:51 Hyratel landing has a lot more variables going on through the entire process
14:52 Hyratel Boostback is very hard
14:53 Hyratel you have to turn the Booster around so it faces Antivelocity, then burn to kill and then reverse Suborbital velocity
14:54 Hyratel then once you do THAT, you have to turn around AGAIN so you're pointed engine-first
14:54 deshipu except they don't have orbital velocity
14:54 Hyratel the BO booster has deployable fins and airbrakes at the top
14:54 deshipu because they never get to the orbit
14:54 Hyratel deshipu,
14:54 Hyratel neither does SpaceX F9
14:55 Hyratel they get goin maybe 1.5km/s
14:55 Hyratel orbital velocity is 7km/s or thereabouts
14:55 deshipu so all this talk about "going to space" is just a pile of marketing
14:55 Hyratel that's why the second stage burns for much longer
14:55 deshipu more like almost 8km/s
14:55 Hyratel space is fuzzy
14:55 Hyratel I mean REALLY fuzzy
14:55 Hyratel there's still measurable aerodynamic drag at geostationary
14:56 deshipu yup
14:56 Hyratel but even suborbital, you're above 99% of the density mass of the atmosphere
14:57 Hyratel (in KSP, the 'atmo cutoff' is 70km, making the atmo curve 60% Earth's
14:57 Hyratel but even the ISS uses its solar arrays as aerosurfaces
14:57 Hyratel so yes. it's *mostly* marketting
14:58 Hyratel but the difference is sufficiently subtle that it can be handwaved and only pedants will get in a snit about it
14:58 Hyratel 'the edge of space' goes from ... 50km altitude up to about 200km
14:58 deshipu also this "experience 0g" they write about is something you can as well get in a falling elevator
14:59 Hyratel but not for four minutes straight
14:59 deshipu because it's not the 0g due to cancelling gravity with speed
14:59 SpeedEvil Altitide is meaningless
14:59 Hyratel even Vomit Comet only had 30s of freefall
14:59 SpeedEvil velocity is important.
14:59 deshipu SpeedEvil: technically, you could move so high that gravity becomes so weak that you can mostly ignore it
15:00 deshipu SpeedEvil: and you could do that at slow speed as well
15:00 Hyratel that said, the BO landing is no less impressive
15:00 SpeedEvil deshipu: yes.
15:00 deshipu of course you would be on an orbit intersecting earth then
15:00 Hyratel deshipu, and it's once you hit atmospheric interface that the hard part of landing comes in
15:00 SpeedEvil deshipu: nope.
15:00 SpeedEvil deshipu: In principle you can escape earth at 1m/s
15:00 SpeedEvil it's just going to take a hell of a lot more fuel
15:01 Hyratel deshipu, so it wouldn't matter if you're coming back down from suborbital or transorbital velocity for your descent flight control
15:01 Hyratel you just have to burn off a lot more speed at turnaround
15:01 Hyratel which does effect the rocket design
15:02 Hyratel but a *lander* is rather different than a Booster
15:02 deshipu I always thought they burn most of the speed with atmospheric friction
15:02 deshipu so you don't have to carry so much fuel
15:02 Hyratel a boostback Booster landing is one of the hardest landings, controls, internal logistics and hardware wise
15:03 Hyratel with a Lander you can have Deep Throttle engines
15:03 Hyratel the Dragon 2 is going to be a powered landing (with backup parachutes)
15:03 deshipu also once you are in the atmosphere, you could have normal jets?
15:03 Hyratel why?
15:03 Hyratel that's added weight and complexity
15:03 Hyratel "the tyranny of the rocket equation"
15:03 deshipu compared to rocket fuel?
15:04 deshipu jets are much more efficient, because you don't need reaction mass
15:04 Hyratel also you have to make sure you can light them and keep them from shredding themselves
15:04 Hyratel yes you do
15:04 Hyratel but you take 80% of your remass from the intake air
15:04 deshipu well, you use the air for that
15:04 deshipu right
15:05 Hyratel but you have to have all the controls and mechanical infrastructure for two types of engines
15:05 Hyratel you've already got the Main Ascent engine
15:05 Hyratel it's Engineering Cheaper to reuse that for the descent
15:06 Hyratel also, jets have a much longer spooling time than a pressure-fed rocket
15:06 Hyratel a pressure fed rocket you can use proportional valves for, and it's even feasible to make a throttlable Turbopump at smaller sizes (less inertia)
15:07 Hyratel yes, it is much less fuel efficient than a jet engine, but the engineering cost is also far lower
15:08 deshipu you pay that once
15:08 Hyratel no, you pay it every time
15:08 deshipu only if you build a new rocket every time
15:08 Hyratel there's a hell of a lot fewer moving parts in a rocket
15:08 Hyratel you have to recert the thrust bits every time, especially if it was a hard landing
15:34 robertj hey all. anyone here familiar with the fancy-pants robotic vacuumes? I've never had one before and am tempted but my down-stairs is ~ 2500sqft and has lots of small interconnected rooms and i'm thinking it's gonna be fail
15:36 robotustra use neato
15:36 robotustra it will not fail
15:36 robertj robotustra, I've heard it gets stuck a bunch
15:37 robotustra I use one - no problem
15:37 robertj how much surface area you vacuuming and which model you got?
15:38 deshipu I have a roomba, it's cool because you have less dust, but it doesn't replace manual cleaning
15:39 deshipu you can do it less often, though
15:41 robotustra 800
15:42 robotustra but it returns to the base to charge
15:42 robotustra as manu times as needed
15:42 robotustra many*
15:42 Hyratel deshipu, it's a smaller frequent upkeep to defer the larger major cleanings?
15:43 deshipu Hyratel: right
15:43 robotustra robertj: usully it charges 1 time for this area
15:43 robertj robotustra, will it be able to find its way through several rooms to a new area?
15:43 deshipu Hyratel: it's mostly automatic
15:43 robotustra for 2500 - buy 2 and split the job
15:44 robotustra easy - it has lidar
15:44 robertj for example laundry -> hall -> kitchen -> dining -> music room?
15:44 robotustra it remember the map
15:44 deshipu robotustra: depends on the model
15:44 robotustra my finds the way
15:45 deshipu mine just bounces off furniture randomly
15:45 deshipu but it was cheap
15:45 robotustra neate signature pro
15:45 deshipu I have a small apartment, so I don't mind
15:45 robotustra oh there is a discount now
15:45 robotustra for them
15:46 robotustra http://www.amazon.com/Neato-Signature-Allergy-Vacuum-Cleaner/dp/B00CBW63QU
15:46 robotustra rumba is as clever as a pingpong ball
15:47 robotustra robertj: I want your house
15:48 robertj robotustra, move to nowhere-USA, it's affordable!
15:48 robertj I thought the XV was supposed to be cheaper than the D8x series
15:49 anonnumberanon I have a Neato but I bought it broken to take the Lidar in it.
15:51 robertj it bugs me that all these guys have navigation but none of them have ways to box out areas as nog-os
15:54 Hyratel there used to be one that came with laser-line barricades
15:58 robertj yeah but that's stupid
15:58 robertj this is 2015 and cameras cost like...$3
15:58 robertj some of these guys are $900
15:59 robertj the D80 is $450 @ BestBuy. They have a 4-year warranty for $70
16:00 Hyratel make sure it's one youc an poke around the programming for
16:00 robertj I've heard Roombas need replacing every 2-3 years, is that still true?
16:00 Hyratel also geospatial awareness is Hard
16:00 Hyratel a line sensor is easy
16:00 robertj Hyratel, yeah but it's $900
16:00 Hyratel so you use it like an electric fence
16:01 robertj maybe i need an out-dated roomba to play with too
16:07 Hyratel early versions are probably pretty cheap
16:08 Hyratel and you could overhaul the control logic for less than a new one
17:12 anonnumberanon NAO's walking speed makes me cry.
17:16 anonnumberanon oh actually it's not too bad if you set it to the fastest speed
17:27 anonnumberanon https://youtu.be/P-J31vyb3P0
17:27 anonnumberanon like such
17:44 jhylands see, now that is a classic shuffle walk
18:12 anonnumberanon its probably slower than some of the gaitsnin robo-one
18:19 anonnumberanon a lot slower than this one for example but this one doesnt have the same range of motion
18:21 anonnumberanon https://youtu.be/-YQIK9e1fks
18:24 anonnumberanon hey what was the neural networks series to watch?
19:00 deshipu ugh neural networks
19:00 deshipu yet another escape hatch
19:05 akem next robo one February 13, anonnumberanon you entering the competition?
19:06 akem "death to humans!" i can here that straight from the future just like jeanne dark
19:06 deshipu kill all humans, for a better future
19:07 deshipu my problem with neural networks is that they are oftern considered as some kind of black box silved bullet by beginner roboticists
19:08 deshipu as in "I have no idea how to do that, so I will just make a neural network and it will figure it out for me"
19:09 deshipu whereas in practice solving a problem with a neural network often involves much more work and deep understanding of the problem than solving it with conventional means
19:10 deshipu it's the typical "this is too hard, instead of solving it I will just make it 10 times harder and delay thinking about it"
19:11 akem they are supposedly good for optical char recogn, but i never really got in touch with that, you know handwriting
19:12 deshipu yes, that's actually well done with a hand-crafted 4-layer network
19:13 deshipu at least for the digits
19:14 deshipu then again, you can also do it using normal template matching or geometrical analysis algorithms
19:18 deshipu eitehr way you need to know what you are doing
19:22 akem deshipu you should try to make your robot spiders learn how to run(and optionnaly how to envennimate a human) by themselves using neural nets
19:23 akem that or genetic algorithms, choose your destiny.
19:23 SpeedEvil akem: https://www.youtube.com/watch?v=OPsSPc-Q5hc
19:25 deshipu akem: that has been done, it's not pretty
19:25 deshipu akem: I'd rather understand how they work
19:26 akem :) speedevil i feel like im at the UAC
19:30 akem robo one website is almost full japanese
19:31 akem possibly i would need 2 japaneses girlfriends to read it correctly
19:33 deshipu that doesn't work
19:33 deshipu you only learn the names for parts of body that way
19:34 deshipu and possibly pieces of clothing
19:37 akem he's comming next robo one http://tinyurl.com/j8m2xa8 get shiny
19:38 akem deshipu it is important, i just need a karaoke version of the website
19:38 akem pieces of clothes i know
19:46 Casper standard cheap screw don't tig well at all... it just go poof
19:50 SpeedEvil https://www.youtube.com/watch?v=O5bTbVbe4e4 - Spacex launch in several minutes
19:58 deshipu it's like watching f1 racing
19:58 jhylands heh
19:59 deshipu everybody just hope for a crash
19:59 ace4016 hah
20:04 mrdata- armyofevilrobots, was it you who did the nasa sample return challenge?
20:41 anonnumberanon Man that shit was so cash...
21:38 robotustra_ scematic to solder http://i.imgur.com/z9LiYPA.jpg
21:58 rue_house wtf
21:58 rue_house why so many votlage changes?
22:14 robotustra_ only 2 steps actually
22:14 robotustra_ 24->12 and 12->5
22:14 robotustra_ some of sensors have 3.3V I'll do 12->3.3
22:15 robotustra_ main bus is 12V
23:11 rue_house so, is it a wheeled robot?
23:16 Jak_o_Shadows This the tity one? I like it
23:26 rue_house "grappling with the most challanging of design problems, the gripper."