#robotics | Logs for 2016-05-23

Back
[02:13:09] <Jinxit> anniepoo_: can't seem to find it
[03:17:23] <rue_house> https://www.youtube.com/watch?v=QNA4icYO7cw
[03:17:26] <rue_house> there is a good one!
[05:41:38] <veverak> that cable hell
[05:41:41] <veverak> :)
[05:59:09] <veverak> http://spectrum.ieee.org/cars-that-think/transportation/self-driving/autonomous-mini-rally-car-teaches-itself-to-powerslide
[07:12:29] <Jinxit> met that guy the other day, doing some really cool stuff
[07:12:38] <Jinxit> and all open-source
[07:40:13] <veverak> damn it
[07:40:18] <veverak> numpy.interp is too slow
[07:40:40] <veverak> numpy.interp( now, (self.start_time, self.end_time), (self.start_pos[i], self.end_pos[i]) )
[07:40:48] <veverak> but, I calculate it on simple [start,end] ranges
[07:40:59] <veverak> maybe there is something better fit for my use case
[07:41:04] <veverak> anybody any tips?
[07:42:18] <Jinxit> I don't see why it would be slow, it's just a few arithmetic ops
[07:42:23] <Jinxit> have you profiled?
[07:44:07] <veverak> yep
[07:44:09] <veverak> :)
[07:44:12] <veverak> that's why I targeted this
[07:44:21] <Jinxit> you doing it to an array or singular values?
[07:44:32] <veverak> singular values
[07:44:45] <veverak> and I don't really have a space to use arrays here I think
[07:44:58] <Jinxit> but you have space to use python?
[07:45:12] <veverak> I thought it will be good enough :)
[07:49:13] <kenalex> hello everyone
[07:50:40] <kenalex> what is a good book on 2D graphics programming?
[07:50:47] <kenalex> eg. scalling a occupancy grid when drawing to screen, rotating 2D coordinates ,etc...
[07:51:36] <z64555> kenalex: I've been going by this: https://en.wikibooks.org/wiki/OpenGL_Programming#Introduction
[07:52:09] <z64555> seems to cover the basics fairly well, there's a section on scientific data and font rendering, too
[07:52:11] <kenalex> thanks
[07:52:43] <kenalex> do you know of any occupany implemntation that I can use to understand how to draw it to the screen ?
[07:53:08] <z64555> yep. they're tutorials instead of just literature
[07:53:59] <kenalex> *occupancy grid
[07:54:28] <kenalex> are familiar with occupancy grids ?
[07:56:12] <z64555> hm, those tutorials don't cover it. but it should be simple enough to create an algortihm to make it
[07:57:33] <z64555> so, you could draw the underlaying grid, and fill in each occupied cell with a square
[07:57:40] <kenalex> i have create one in memory. just having some issues trying to visualize it by drawing it to the screen
[07:58:36] <z64555> ok, do you have the individual cells, or do you still have geometry in there
[07:59:29] <kenalex> individual cells
[08:00:05] <z64555> then you've pretty much got it already
[08:02:56] <kenalex> ok
[08:03:06] <z64555> you can use GL_QUADS to make things a bit easier
[08:03:40] <z64555> it tells the GPU to draw an individual quad (square in your instance) for every 4 vertices you pass it
[08:04:05] <kenalex> my main convern is that the scrren has limited size say 800 x 600 and I want to plot the points that are in catesian coordinates to screen coordinates
[08:05:14] <Jinxit> I recommend using something that's not raw GL for this stuff
[08:05:41] <z64555> no problem, you can use a uniform variable in the vertex shader to scale up/down things
[08:06:24] <z64555> hm, I think there's a plotting program available on linux machines
[08:06:46] <z64555> saw some mention of it. but I'm a blacksheep of IRC, I'm a windows user.
[08:06:58] <kenalex> ok
[08:07:20] <kenalex> I am manily interested in doing the drawing implementation using opengl
[08:12:01] <z64555> the toughest thing I've found about opengl is the frustration you have to go through to get your build environment set up
[08:13:17] <kenalex> done that using a 3rd party library SFML
[08:13:30] <z64555> the tutorials I linked use SDL and glew, SDL was ok, but I had to build glew. Thankfully there was a .sln in there already so I didn't have to fight with it
[08:19:11] <z64555> SFML looks like a good API, I'm not sure how transparent it is to the underlaying LGL code, however
[08:20:17] <z64555> oh boo, they've deprecated Quads
[08:24:05] <z64555> ok. looks like it covers the basics fairly well. I'm not too fond of their "OOP everything!" approach
[09:55:27] <TheFactory7> veverak: wow that car is impressive. I'd be very keen to know if they are using both the gpu and the cpu to its fullest potential or whether one is heavily underutilized (i suspect the cpu)
[10:15:56] <Jinxit> ask them :)
[10:25:54] <veverak> TheFactory7: yep, ask them
[10:27:17] <TheFactory7> i shall. seems like a great project and sweet kudoz for it being open. Are those kind of cameras expensive as far as you know?
[10:30:35] <veverak> dunno
[10:30:37] <veverak> but, in this project
[10:30:45] <veverak> I am little bit worried about /watt effeciency
[10:30:47] <veverak> :)
[10:45:09] <TheFactory7> can you elaborate?
[10:45:40] <TheFactory7> veverak: overpowered in terms of motors or in terms of controller/hardware
[10:53:58] <veverak> that GPU is not exactly power effecient
[10:54:06] <veverak> computing_power/watt effeciency
[10:54:08] <veverak> :)
[14:35:26] <Donitzo> what do you call the alternative to a latch relay, a relay which doesn't remember its position?
[14:35:43] <SpeedEvil> normal
[14:35:50] <Donitzo> normal relay, okay that works
[14:35:58] <SpeedEvil> you don't specify it as it's assumed
[14:36:01] <SpeedEvil> so just 'relay'
[14:54:29] <veverak> YEAH
[14:54:32] <veverak> working genetic algorithm
[14:55:01] <veverak> deshipu: my genetic algorithm found out that ideal leg timing is 0.25/0.75/0.25/0.75
[14:55:05] <SpeedEvil> yay
[14:55:26] <veverak> and evaluation is written in a way that it won't allow 0.25/0.25/0.5/0.5
[14:55:28] <veverak> :)
[15:04:07] <Jinxit> why not gradient descent instead of evolutionary?
[15:05:19] <veverak> not sure if aplicable porperly
[15:05:50] <SpeedEvil> I loved the paper on genetic evolution of gait and muscle attach at the same time
[15:07:27] <veverak> well, it's true that genetic algorithm got high 'cool' level
[15:07:39] <veverak> but I think I will need to play a lot with it to be able to use it practically
[15:08:34] <Jinxit> if there's a loss and a differentiable function to calculate that loss, then there's a gradient
[15:08:43] <Jinxit> but that depends on how you do the locomotion I suppose
[15:09:17] <veverak> Jinxit: walking of quadruped robot
[15:10:14] <doomlord> loss in this case, -distance moved / energy expended?
[15:10:35] <veverak> evaluation -> use parameters to simulate walking of robot
[15:10:44] <veverak> fitness -> time it takes for robot to walk specific distance
[15:12:47] <doomlord> is this impossible for gradient descent.. you're not really correcting the output?
[15:13:04] <doomlord> the simulation (to calculate loss) is not actually differentiable?
[15:14:15] <veverak> I can't really imagine how to be honest :) or I don't understand gradient descent properly
[15:14:41] <veverak> whatever, task definition is to use genetic algorithm anyway
[15:14:43] <veverak> :)
[15:25:05] <[VEGETA]> is there a way to connect between ros and freertos?
[16:09:56] <[VEGETA]> ...
[16:28:25] <veverak> fuuuu
[16:28:29] <veverak> lambdas in python are evil
[16:35:06] rue_house changed topic of #robotics to: If you ask a question, you must wait for the answer. | http://tom-itx.no-ip.biz:81/~webpage/index.php | D is for Droid
[16:37:21] <LoRez> veverak: how so?
[16:38:55] <veverak> some unexpected behavior
[16:38:57] <veverak> :)
[16:39:16] * veverak sort of thought that variables used from enviroment where lambda is created are 'catched' during lambda creation
[16:39:21] <veverak> but, given that all is reference
[16:39:24] <veverak> they were not
[18:07:20] <rue_shop5> well, nto that I had to, but making a spindle for the new cnc isn't comming out that bad
[18:22:30] <TheFactory7> have you seen the xiaomi ninebot mini? it doubles as transport as well as an open (?) and modula robotic platform (i.e. video screen head with cameras and arms)
[18:24:36] <TheFactory7> doesn't seem on sale yet
[18:25:28] <theBear> that's a lotta weird words... did you just make all that up, or does it make real sense ?
[18:26:21] * theBear pretends to have a video screen head and cameras for arms and does his finest pop+lock combo, followed by a short, understated robot moves finish
[18:39:40] <TheFactory7> theBear: this dude http://uncrate.com/p/2016/01/segway-robot.jpg
[18:40:15] <theBear> lol, issa so cute
[18:40:21] <TheFactory7> the arms are a bit toyish but you can probably mount something more interesting on
[18:40:32] <theBear> can you stand on his err, hips and ride along ?
[18:40:52] <TheFactory7> yep, 30 km range, i think 20km an hour or so
[18:41:12] <rue_shop5> you know, if you were standing on it, its hands and face would be right about at the same height as your...
[18:41:36] <TheFactory7> the head goes back inside (moves 180 degrees)
[18:41:40] <TheFactory7> and the arms can be removed fully
[18:42:30] <TheFactory7> you use the knees to turn, apparently much more stable than the 2015 scooter thing
[18:42:49] <rue_shop5> mhm
[18:43:21] <TheFactory7> these ones to be clear http://img.banggood.com/thumb/view/2014/chenjiawei/12/SKU188052/SKU18805201.jpg
[18:44:37] * rue_shop5 is suffering from "I'v done something, I want to sleep now"
[18:45:12] <TheFactory7> today i also learned that the info entertainment system in my car is in JS .. :(
[18:45:31] <Tom_itx> rue_shop5, do something more
[18:45:51] <Tom_itx> what did you make on your 4jaw lathe?
[18:46:04] <rue_shop5> hard when there isn't the mental energy to get out of the chair
[18:46:29] <Tom_itx> bring the work to you
[18:46:35] <rue_shop5> I rebored a sync pulley for the cnc cutter and made a shaft for it too
[18:47:19] <Tom_itx> i wanna see results from that cnc soon
[18:47:52] <rue_shop5> heh
[18:48:10] <rue_shop5> I have a project that involves 3 holes in a peice of wood
[18:48:13] <rue_shop5> er, 4
[18:48:20] <rue_shop5> I think I can hand code that one...
[18:50:38] <Snert_> I used Drill v1.2 for that one.
[20:08:42] <rue_shop5> yes, but they have to be in the right spot
[20:08:51] <rue_shop5> and If I dont have to drill them all the beter
[20:09:14] <rue_shop5> set up a few jobs on the table, work out all the home locations
[20:09:23] <Tom_itx> http://tom-itx.no-ip.biz:81/~webpage/cnc/temp/Kurtvise2.jpg
[23:37:18] <rue_shop5> haha just did some hand coded cutting with the cnc
[23:37:22] <rue_shop5> the machine sucks!
[23:39:39] <z64555> how so?
[23:49:05] <rue_shop5> too flimsy, power supply keeps kicking out...
[23:49:24] <rue_shop5> its bigger and faster than my old one, but thats it
[23:49:32] <rue_shop5> I'm gonna make a better one later
[23:49:45] <rue_shop5> getting this one going got a lot of ground work set
[23:57:54] <z64555> bigger buffer cap, maybe?
[23:57:58] <z64555> for the power supply