#robotics Logs
Apr 17 2017
#robotics Calendar
12:01 AM rue_house: ok I updated it
12:06 AM rue_house: http://ruemohr.org/~ircjunk/robots/arm10/p1080879.jpg
12:06 AM rue_house: http://ruemohr.org/~ircjunk/robots/arm10/p1080872.jpg
12:06 AM rue_house: I might get it all wired up tommorow
12:06 AM rue_house: maybe not, I been up till 1am for the last 3 days, and tommorow I have up be up AT 6am
12:06 AM rue_house: its gonna be a painfull day
12:06 AM gottaname|wurk: awww
12:06 AM gottaname|wurk: well, sleep is a luxury
12:10 AM rue_house: getting luxury served to you while your driving downt eh highway is not appropriate
12:11 AM rue_house: http://ruemohr.org/~ircjunk/robots/arm10/p1080863.jpg
12:11 AM rue_house: thats the 4th of that arm design
12:11 AM rue_house: I really need to work out soemthing for them to do
12:11 AM rue_house: today I was supposed to [lots of things]
12:12 AM rue_house: but one of them was to work out that servo force feedback I been going on about
12:12 AM gottaname|wurk: rue_house, if you attach a dildo to it?
12:12 AM gottaname|wurk: :P
12:12 AM rue_house: so, appologies, maybe on tuesday
12:12 AM * rue_house beats Gottaname over the head with a pillow
12:12 AM rue_house: SLEEP SLEEEEEEP
12:13 AM gottaname|wurk: lies
12:13 AM gottaname|wurk: no sleep
12:24 AM rue_house: esp while driving
01:26 AM z64555: sleep now, die later (not sooner)
01:54 AM gottaname|wurk: pretty quiet
03:03 AM Jak_o_Shadows: aye
03:14 AM Anniepoo_: jay
03:14 AM Anniepoo_: i'm learning some blender game engine for snail
03:15 AM Jak_o_Shadows: I never figured out how to move bones programmatically
03:16 AM Anniepoo_: set up drivers
03:22 AM Jak_o_Shadows: ?
03:22 AM Jak_o_Shadows: Is that game logic blocks?
03:22 AM Jak_o_Shadows: Or putting in keypoints before the GE?
03:25 AM gottaname|wurk: ewww
03:25 AM gottaname|wurk: blender game engine
06:58 AM Jak_o_Shadows: /join #upbgeusers
09:06 AM jandor: how goes it?
10:21 AM cnnx: why is this motor so expensive? i'm looking for an outdoor robot dc motor to carry around 50-100lbs
10:21 AM cnnx: http://www.globalindustrial.ca/p/motors/ac-motors-2-phase/dc-motors/motor-xs56c-tefc-1800rpm-c4d17fk28c?infoParam.campaignId=C5E&gclid=CjwKEAjwz9HHBRDbopLGh-afzB4SJABY52oFLdxrpe61OuikmhyDoh5bWlhHvjiXf4lNSGpICNdXwhoCipPw_wcB
10:21 AM cnnx: 4 of these is like over 2k
10:24 AM SpeedEvil: Because it's properly specified and reliable.
10:25 AM cnnx: is there an online calculator that will help me choose the right size specced DC motor for my application?
10:25 AM cnnx: i think 1HP 12VDC might be overkill
10:26 AM SpeedEvil: Start out with a basic understanding of physics.
10:26 AM Snert: set a budget limit. Throw out all motors that are over budget.
10:26 AM cnnx: ok
10:26 AM SpeedEvil: How many horses would it need to drag your robot.
10:26 AM cnnx: not even 1
10:26 AM cnnx: i can drag a robot of 100lbs with wheels
10:27 AM SpeedEvil: Are you a horse? You never know on the internet.
10:27 AM cnnx: no
10:28 AM SpeedEvil: Look at rating of similar things
10:28 AM SpeedEvil: http://www.ebay.co.uk/itm/New-800W-Battery-Kids-Quad-ATV-Disc-Brakes-Suspension-Front-Rear-/291960505556?hash=item43fa33c4d4:g:TJMAAOSw44BYQXRz
10:28 AM SpeedEvil: youtube, and see what it can do, for example
10:28 AM cnnx: ok
10:30 AM cnnx: SpeedEvil: http://www.robotshop.com/blog/en/drive-motor-sizing-tool-9698
10:38 AM SpeedEvil: The above is misleading in some ways, you want to oversize your motors over that, unless everything is absolutely perfect
11:11 AM jandor: what about inertia matching?
11:39 AM Gottaname is now known as Logikoma
12:18 PM anniepoo: ok, now I'm convinced to use YARP
12:18 PM anniepoo: the logging message that the nameserver's up is
12:18 PM anniepoo: Name server running happily
01:11 PM mumptai_: re
01:12 PM anniepoo: bop de bop
01:12 PM mumptai_: what is yarp?
01:13 PM mumptai_: http://www.yarp.it ?
01:13 PM anniepoo: yes
01:13 PM anniepoo: it's like ROS, but not as invasive
01:14 PM mumptai_: ooookay? i was under the impression that ros had it problems, but was the only thing worth adopting (if one has no legacy ties) atm
01:16 PM mumptai_: but we kinda avoided it altogether, because the pc-based things have been to brittle for use in the robotics competitions
01:16 PM mumptai_: also realtime and low-level IO is kinda tricky with PC stuff and also embedeed linux
01:18 PM anniepoo: it's winning because it's winning because it's winning?
01:19 PM anniepoo: YARP and a few others basically arose out of reaction to the awkwardness of working with ROS
01:20 PM mumptai_: we used 8 and 32bit microcontrollers, some FPGAs and CAN. but stuff like full logging and image-processing are beyond this type of system
01:22 PM anniepoo: right
01:23 PM anniepoo: well, your edge devices aren't where you do such things
01:23 PM mumptai_: depends
01:23 PM mumptai_: actually logging at raw sensor level isn't a luxury in all cases
01:24 PM anniepoo: YARP has plug in "carriers" structure
01:24 PM mumptai_: image proceesing is currently being word on, and will be a rpi3 that will be a blackbox outputting locations on detected stuff in view of the camera
01:25 PM anniepoo: so you can talk from your device via something that works for it
01:25 PM anniepoo: (eg usb/rs-488 emulation for an arduino)
01:25 PM anniepoo: lots of carriers already implemented
01:34 PM solol: what's up
01:40 PM anniepoo: hey solol
01:40 PM solol: how's the robotic world these days?
01:40 PM solol: people making strides in lightening things up?
01:40 PM solol: making things cheaper?
01:40 PM * anniepoo is here, being chased around the world by the robot from Old Glory Insurance commercial
01:41 PM anniepoo: s/world/room/
01:43 PM solol: haha
01:45 PM solol: not familiar with the robot.. bout to youtube that
01:45 PM solol: SNL?
01:45 PM anniepoo: classic - SNL skit
01:46 PM anniepoo: predatory commercial targeting elderly people
01:46 PM anniepoo: definitely look it up- every roboticist should know it!
01:47 PM solol: 1995
01:49 PM solol: lol @ they eat old peoples' medicine for fuel
01:52 PM anniepoo: 8cD
01:53 PM solol: are people getting good at designing the AI for robot sensors and stuff?
01:53 PM anniepoo: it's slow steady progress
01:53 PM anniepoo: no amazing breakthroughs
01:54 PM anniepoo: ML is getting good, and that's helping
01:54 PM mumptai_: visual slam got cool
01:54 PM anniepoo: yes,that's working
01:55 PM solol: what's ML
01:55 PM solol: or who
01:55 PM anniepoo: machine learning
01:55 PM solol: yeah ok
01:55 PM solol: what problems are they solving?
01:55 PM solol: or have solved
01:56 PM mumptai_: https://www.youtube.com/watch?v=GnuQzP3gty4
01:59 PM solol: that's sweet
02:29 PM solol: is automation really a thing that will happen to 50% of the jobs in the next 10-15 years?
02:30 PM solol: i've never seen really well functioning robots
02:30 PM solol: i know the auto industry has some though
02:34 PM deshipu: pick-and-place machines work pretty well too
02:36 PM solol: what are those used for besides electronics?
02:36 PM deshipu: electronics
02:37 PM solol: what do they do for smaller electronic parts
02:37 PM solol: like cpus
02:37 PM deshipu: robots
02:39 PM weyland|yutani: solol:every invention throws some kind of jobs under the bus
02:39 PM solol: yeah hopefully anyway
02:40 PM solol: people should be better off than they are in some ways
02:40 PM deshipu: we live longer
02:40 PM solol: and the world should be better off than it is because of humans
02:40 PM deshipu: pharmacy uses a lot of robots too, btw
02:41 PM solol: what's a pharmacy use a robot for
02:41 PM deshipu: also all those blood sample and tissue testing is mostly automated nowadays
02:41 PM solol: pharmacies dont do much
02:41 PM deshipu: oh really
02:41 PM deshipu: maybe have a search on youtube
02:41 PM weyland|yutani: we are better off 150 years ago you probably would work 12 hours a day 6 days a week in some horrible industrial setting with machines so bad designed using them is like juggling with chainsaws
02:41 PM solol: i thought they just distribute pills
02:41 PM deshipu: solol: think better :)
02:41 PM solol: yeah people weren't expected to live past 35 back 200 years ago
02:42 PM solol: what do you mean by think better
02:42 PM solol: until antibiotics were invented/discovered anyway
02:43 PM deshipu: solol: right now you are making pretty dumb assumptions which can be easily updated with a simple youtube or wikipedia search
02:43 PM solol: yeah but pharmacies basically just count pills
02:43 PM solol: unles you really dont want a pharmacist you can pretty much say that job is safe
02:43 PM deshipu: and where do those pills come from?
02:44 PM solol: automation is an old thing in chemistry from what im told
02:44 PM deshipu: yup
02:44 PM deshipu: and works very well there
02:44 PM deshipu: also all the processed foods industry
02:45 PM solol: yep
02:45 PM solol: so what is next?
02:45 PM deshipu: you think that frozen pizza was made by a human?
02:45 PM weyland|yutani: solol: pharmacies nowadays have fully automated picking machines they just scan in your reciept and a automated machines searches its storage area and goves out the right pills
02:45 PM solol: that's kinda neat
02:45 PM deshipu: by "pharmacy" I meant the industry, not the outlets that sell pills
02:46 PM solol: lol ok
02:46 PM weyland|yutani: i meant the outlets
02:46 PM deshipu: sorry for not making it clear
02:46 PM solol: i kinda meant what is next for automation
02:46 PM deshipu: weapons industry is heavily automated too
02:46 PM weyland|yutani: taxi trucks cars
02:46 PM deshipu: they are pushing hard for automation in agriculture
02:46 PM solol: yeah that's a big industry
02:47 PM deshipu: also stuff like cutting trees or making building materials in quarries
02:47 PM solol: agriculture would be nice to automate... there's just a thing like deer that could cause problems, but they already can steer with gps
02:48 PM solol: cutting trees would seem kinda hard to automate
02:48 PM anniepoo: solol, I live in a forestry area
02:48 PM deshipu: it's not fully automated, but a lot of automation is there already
02:48 PM anniepoo: it's already highly automated
02:48 PM anniepoo: truck with an arm grabs tree
02:49 PM solol: i guess im thinking more trees near houses or in your yard
02:49 PM solol: where you gotta watch where they fall
02:49 PM solol: do they haul the tree out of the woods upright?
02:49 PM anniepoo: you have to watch where they fall in the forest too
02:49 PM solol: oh alright
02:50 PM anniepoo: no, the grabber machine places it on the ground. Then cutters on the 'hand' lop off branches
02:50 PM weyland|yutani: solol http://gpsworld.com/machine-control-agprecision-agsensor-fusion-forestry-10146/
02:50 PM anniepoo: then hand puts it on truck
02:50 PM anniepoo: it's not fully automated, but it's highly mechanized
02:50 PM anniepoo: once it gets to the sawmill, there's a lot of automation
02:50 PM deshipu: there are also exteremely cool robot-trains for building and re-building train rails
02:51 PM solol: nice
02:51 PM solol: so a lot of big things have happened
02:51 PM anniepoo: and we're installing robotic machining tools right now
02:51 PM anniepoo: as in 'lots of dust in that room, big truck will arrive in next couple weeks' right now
02:52 PM anniepoo: and they aren't our first
02:52 PM solol: cool
02:52 PM anniepoo: I'm running an automated laser cutter right now, as in
02:52 PM anniepoo: (2 mins, have to change the sheet in it)
02:52 PM solol: what are the limits they are trying to overcome right now?
02:53 PM deshipu: "they"?
02:53 PM solol: machine learning was mentioned
02:53 PM solol: engineers
02:53 PM deshipu: all of them :)
02:54 PM anniepoo: back
02:54 PM solol: that was quick
02:54 PM anniepoo: deshipu++
02:54 PM anniepoo: yes - I'm making a fiberglass shell for a robotic snail
02:54 PM anniepoo: I modeled it in 3D software
02:54 PM weyland|yutani: solol:just look at the last darpa challenge...boston dynamics robot did not walk very good on 2 feet back then and 6 months lather they released a video of atlas walking in uneven snowy woodland terrain
02:55 PM solol: yeah that's what im talkin about
02:55 PM anniepoo: then used other software to cut it into slices
02:55 PM anniepoo: I'm importing the slices into laser control software
02:55 PM deshipu: weyland|yutani: to be honest, the DARPA challenge is a bit contrieved example
02:55 PM anniepoo: loading a piece of corrugated cardboard on the laser
02:55 PM anniepoo: and pressing start
02:56 PM solol: doesn't cardboard catch on fire
02:56 PM deshipu: weyland|yutani: they are trying to come up with a situation that would justify a humanoid robot -- which is stupid
02:56 PM anniepoo: 4-5 mins later it beeps at me
02:56 PM anniepoo: nope
02:56 PM weyland|yutani: deshipu, why is a bipedal robot stupid?
02:56 PM anniepoo: I take the cardboard slices out - usually a half dozen of them
02:57 PM deshipu: weyland|yutani: any other configuration will be more efficient for pretty much any task
02:57 PM anniepoo: then I sweep out the little circles left when it cut out the alignment pin holes
02:57 PM anniepoo: put in new cardboard
02:57 PM deshipu: weyland|yutani: even in the darpa challenge, the robots that had wheels fared the best
02:57 PM anniepoo: come back and repeat
02:58 PM anniepoo: I'm on 2, when I get to 12 I take the pieces, a jig with steel rods, laminate pieces
02:58 PM anniepoo: cover shell with saran wrap and then plaster bandages. a layer of plaster,
02:58 PM solol: can fiberglass be made to be pretty strong?
02:59 PM solol: i've only seen that it can be made durable
02:59 PM solol: which is kinda meaning not strong
02:59 PM anniepoo: take that off shell, then hand sand, then fiberglass inside it
02:59 PM weyland|yutani: solol, boats are made out of it
02:59 PM z64555: ^
02:59 PM solol: water isn't that hard though
02:59 PM anniepoo: I used ot live on a boat made of it
02:59 PM z64555: ohohohho.
02:59 PM anniepoo: LOL
02:59 PM anniepoo: the forces on ships are HUGE
02:59 PM z64555: water at the speed of several miles per hour is
03:00 PM solol: cool. what is it made out of
03:00 PM anniepoo: water in the form of ocean waves are
03:00 PM solol: what makes it stronger
03:00 PM solol: or durable or whatever
03:00 PM anniepoo: glass fibers and resin
03:00 PM anniepoo: glass is actually really strong
03:00 PM anniepoo: but it's brittle
03:00 PM anniepoo: making it into fibers removes the brittleness
03:01 PM solol: hm ok
03:01 PM anniepoo: most canoes, many car bodies, especially those of fancy custom cars, many airplane parts, the hulls of most yacht sized boats,
03:01 PM weyland|yutani: solol, this itchi yeollow insulation in attics...thats glass too
03:03 PM z64555: glass fibers are significantly more ductile (bendy) than sheets. If you bond a bunch of fibers together in a weave/cloth like fabric and then set them in a water proof resin
03:03 PM z64555: you get fiberglass
03:03 PM solol: yeah i kinda knew the process
03:03 PM solol: i thought it would be the resin that is bendy/weak
03:04 PM solol: but it'd be the weave probably too
03:04 PM z64555: the majority of the strength is from the glass, the resin just holds it together
03:04 PM anniepoo: the resin is weak, but it's not carrying the load
03:04 PM z64555: the weave is what determines how well the load is distributed across the structure
03:04 PM z64555: which, coincidently, is also how kevlar vests work
03:05 PM anniepoo: the space shuttle's main tank was made of fiberglass
03:06 PM SpeedEvil: anniepoo: no, it wasn't.
03:07 PM SpeedEvil: anniepoo: It was made of aluminium, or latterly aluminium/lithium
03:07 PM anniepoo: carbon fiber is one of the strongest available engineering materials, and it's just like fiberglass, except they use carbon fiber
03:07 PM SpeedEvil: the SRBs were steel.
03:07 PM anniepoo: oh, thanks - I was wrong
03:07 PM anniepoo: I did know that
03:08 PM SpeedEvil: SpaceX is proposing direct unlined carbon fibre to contain liquid oxygen and methane in their new vehicle.
03:08 PM z64555: er... I'm still new to the carbon fiber lingo. direct unlined?
03:09 PM anniepoo: I was standing there when that was first tested 8cD
03:09 PM SpeedEvil: carbon-epoxy, without a stainless steel or other liner
03:09 PM jandor: hi again
03:09 PM anniepoo: Will Smith - discovered some process to do it - brought a sample into tech shop and tested it
03:09 PM anniepoo: hey jandor
03:10 PM z64555: hmm... I wonder about temperature differences on the epoxy
03:10 PM anniepoo: and Speed, yes, wikipedia agrees with you
03:10 PM solol: will smith the actor?
03:10 PM solol: discovered unlined carbon fiber?
03:10 PM anniepoo: no
03:10 PM anniepoo: will smith the physicist
03:10 PM anniepoo: not somebody famous
03:11 PM anniepoo: team pheonix, Ansari/Google X prize
03:11 PM anniepoo: the lunar lander
03:11 PM z64555: Lord Nerds, basically
03:11 PM z64555: :)
03:11 PM anniepoo: though oddly what they mostly ended up doing was making rocket motors for other X prize contestants
03:11 PM solol: he discovered unlined carbon fiber though?
03:11 PM z64555: no
03:11 PM anniepoo: yes
03:12 PM z64555: oh
03:12 PM solol: alright
03:12 PM anniepoo: as in process for making carbon fiber that could stand up to LOX
03:13 PM anniepoo: James Erd machined the rocket motors they made. Did lots of prep work at the next bench over from me.
03:14 PM anniepoo: 8cD James also made trees out of cheetos
03:15 PM z64555: lolwat
03:16 PM anniepoo: he made movie props, including a tree made from cheetos
03:17 PM z64555: what did he use to bond them? where they still powdered?
03:17 PM anniepoo: hot glue, I think (honestly, I don't remember). Yes, still powdered.
03:18 PM z64555: huh.
03:18 PM anniepoo: he had to use a full face respirator, turns out cheetos in huge quantities are pretty toxic
03:18 PM solol: lol
03:18 PM solol: so this was a large tree
03:18 PM solol: i was thinking maybe a desktop sized tree
03:18 PM anniepoo: bit larger than desktop
03:19 PM anniepoo: about 7 ft tree
03:19 PM anniepoo: he also made a doghouse out of cheetos
03:19 PM anniepoo: but he made fake cheetos for that
03:19 PM anniepoo: yah... lord nerds
03:20 PM weyland|yutani: some are collectiong ears others buld trees from cheetos
03:20 PM anniepoo: me, I'm making a snail
03:22 PM weyland|yutani: its a robot
03:22 PM anniepoo: it is indeed
03:22 PM anniepoo: 8cD
03:22 PM weyland|yutani: and put lasers on it
03:22 PM jandor: but does it know it?
03:22 PM anniepoo: no, and yes
03:23 PM anniepoo: or 'will'
03:23 PM anniepoo: NLP not built yet
03:23 PM jandor: what is nlp
03:23 PM anniepoo: natural language processing
03:24 PM anniepoo: in general, problems of extracting meaning from unstructured human speech or writing
03:26 PM jandor: ah
03:26 PM jandor: i'm constanly forgot what it is
03:26 PM jandor: i did some in prolog about 10 years ago
03:26 PM anniepoo: for the snail, it will be such tasks as interacting with audience members
03:26 PM jandor: basic processing
03:26 PM anniepoo: 8cD
03:29 PM jandor: I don't know what to do
03:29 PM jandor: I've made one more pull request to ros
03:30 PM jandor: the more I see, more I dislike it
03:30 PM jandor: not the idea, but organization of it
03:30 PM anniepoo: you know, my distaste for ROS is not that I've acutally used it much, or even that I have some fundamental beef with the architecture (though I do)
03:31 PM anniepoo: but that you're FAR from the first person I've heard that from
03:32 PM anniepoo: we have a fundamental trap in computing
03:32 PM weyland|yutani: that is?
03:32 PM anniepoo: whenever an overly complex system, like ROS, or Apache, or whatever, acheives market penetration
03:33 PM anniepoo: a set of programmers struggle and learn the complexities of it. They then become defenders of the status quo
03:33 PM anniepoo: because they have an investment in it
03:34 PM weyland|yutani: anniepoo, do you really think there will be an de facto standard robot os universally used?
03:35 PM solol: hopefully
03:35 PM solol: why wouldnt there be
03:35 PM weyland|yutani: anniepoo, and apart from that its the same with C vs C++ vs java vs c# vs ......
03:35 PM jandor: personally i defend python
03:36 PM jandor: lol
03:36 PM jandor: I'm trying to overteach my friend
03:36 PM jandor: you know
03:36 PM jandor: he's a former asm programmer
03:36 PM solol: i don't really like python
03:36 PM jandor: security, viruses, antiviruses and so on
03:36 PM solol: it had some good things and bad things
03:36 PM jandor: so then he started to write in delphi
03:36 PM solol: i can't say i really disliked it too much
03:36 PM weyland|yutani: jandor, your sounding like he was a drug addict ;P
03:36 PM jandor: may be 15 years ago
03:37 PM jandor: he pretty is asm-addict, yep
03:37 PM jandor: so it was like s1,s2,s3: string; i1,i2,i3: integer
03:37 PM weyland|yutani: "will write drivers for food"
03:37 PM jandor: and using them as registers
03:37 PM jandor: with function names like f1 and fr2
03:38 PM jandor: with r standing for "recursive"
03:38 PM jandor: then he at last learned it , and as you say, did a major investment in it
03:38 PM jandor: in the means of time
03:39 PM jandor: now I insist on switching to python (or other language, but I write in python, so it is python)
03:39 PM jandor: and most of his code in python is delphi-like
03:40 PM jandor: I almost hear his inner gears cracking
03:40 PM jandor: in the head
03:40 PM jandor: then he tries to write pythonic code in python
03:40 PM jandor: so what annie says about time investments is totally true
03:41 PM jandor: and there is no point in arguing python over ruby etc. Only time tells who was right
03:42 PM jandor: the only argument that is valid on my mind is "show me your code"
03:42 PM jandor: and then we can talk about aproaches, methods, languages and so on
03:42 PM jandor: ok, finished complaining
03:42 PM jandor: * GOD I HATE PHP
03:43 PM jandor: ok, now finished really
03:44 PM solol: what did annie say about time investments
03:44 PM solol: i missed it
03:45 PM anniepoo: hi
03:45 PM jandor: sent to private
03:46 PM solol: thx
03:47 PM jandor: at out point we're about starting anonymous delphi user group
03:47 PM jandor: with all that "hi, my name is eugene and I write in delphi"
03:47 PM jandor: "hello eugene"
03:47 PM anniepoo: lol
03:48 PM anniepoo: 8c( I've been there with Java
03:49 PM z64555: I think the C family has a strong foothold because of its relationship with asm, but then again the same could be said for others, like cobol or even Fortran
03:49 PM z64555: ...and then there's VHDL
03:50 PM jandor: hey guys
03:50 PM jandor: and ladies
03:50 PM jandor: what you think about rust?
03:50 PM anniepoo: looks interesting, haven't had time to look in detail
03:50 PM weyland|yutani: well C is used in µC heavily so i dont think it will vanish in obscurity very fast
03:51 PM weyland|yutani: z64555, vhdl is black magic
03:51 PM z64555: lol, ikr?
03:51 PM anniepoo: 8cD CHR is black magic!
03:52 PM jandor: what is chr
03:52 PM anniepoo: Constraint Handling Rules, by Thom Freuwirth
03:52 PM anniepoo: you know prolog, you know how variables in logic languages express knowledge
03:53 PM jandor: ok
03:53 PM anniepoo: but a traditional logic lang only allows for complete knowing, and complete ignorance of an atomic value
03:53 PM jandor: ok
03:53 PM jandor: so here the intuitionists come, aren't they?
03:54 PM solol: ooh, a book i should read?
03:54 PM anniepoo: a constraint language allows applying constraints to a variable
03:54 PM jandor: aha
03:54 PM anniepoo: http://www.pathwayslms.com/swipltuts/clpfd/clpfd.html
03:54 PM jandor: so we can specify how much do we know about it
03:54 PM jandor: right?
03:54 PM anniepoo: yes
03:54 PM jandor: ok, got it
03:54 PM anniepoo: we can specify a domain
03:55 PM anniepoo: and if we specify X #< Y
03:55 PM anniepoo: then say that Y is in the range 1..5
03:55 PM jandor: can it automatically find solutions for linear programming?
03:55 PM jandor: like simplex or smthlike that
03:55 PM anniepoo: then we restrict X's domain by discarding values 5..inf
03:55 PM anniepoo: yes, actually.
03:56 PM anniepoo: I know implementing simplex in it is pretty trivial
03:56 PM anniepoo: but tbh I don't know how to do it
03:56 PM anniepoo: now, CHR is a specialized language for constructing new constraint systems
03:57 PM anniepoo: clp(fd), the thing I linked tutorial for, gives you a fixed set of constraint types
03:57 PM anniepoo: and a hook to create new ones
03:57 PM anniepoo: but CHR is a rule production system for makeing new ones
03:57 PM anniepoo: =8cO it's arcane!
03:58 PM jandor: why
03:58 PM anniepoo: brb - changing the laser's diapers
03:58 PM jandor: rule production should be much alike to prolog's resolution
03:58 PM jandor: or how it is called
04:00 PM anniepoo: prolog is forward chainging
04:00 PM anniepoo: yes, prolog's unification-resolution cycle
04:01 PM waterjar: Hello everyone, I'm using FreeRTOS on a Cortex M3, and for some reason my tasks aren't running: when I step through with a debugger, it's always in the idle task
04:03 PM robopal: is there a simulation tool for that chip? does it work there?
04:04 PM waterjar: robopal: I don't have one up and running
04:04 PM waterjar: I think I'm making a silly mistake somewhere: I'm new to RTOSs
04:05 PM waterjar: robopal: doesn't immediately strike me as something that would be a hardware issue though
04:05 PM mumptai_: because it is statistically more likely to be idling
04:06 PM mumptai_: usually your debugger need support for the used RTOS
04:06 PM mumptai_: otherwise your context is jumping all over the software
04:07 PM Kozuch: Hi there, is monocular visual odometry able to estimate relative scale? I mean say I have a sequence of 10 images that are taken on a track each 1 m after the previous, can some mono odometry method distinguish relative scales when it processes image pairs that are in various distances from each other? I mean like processing 1st vs 10th image and 9th vs 10th image - will the fist give 10x relative scale than the second? I am examining this
04:07 PM Kozuch: OpenCV based odometry code (https://github.com/avisingh599/mono-vo) but it only gives something like "translation vector" that always has a size of 1 regardless the distance measured... I know mono odometry can not do absolute scale but I thought it can do relative (question is what "relative" actually means here)...
04:08 PM waterjar is now known as stephen_
04:09 PM jandor: well
04:09 PM jandor: what is measured?
04:09 PM jandor: what odometry exactly does? I'm not very familiar with it
04:09 PM jandor: is it external camera or is it on the rover?
04:10 PM Kozuch: jandor, odometry measures "distance travelled" from camera images (or video)
04:10 PM jandor: yep, but is that camera external?
04:10 PM Kozuch: what do you mean by "external"? it is usually mounted on the robot directly...
04:10 PM SpeedEvil: Kozuch: you use it as scale based on the robot scale
04:11 PM jandor: aha, that's what I was asking
04:11 PM SpeedEvil: Kozuch: so you know the scale with regards to the camera height from the floor
04:11 PM SpeedEvil: If you double the camera height, the scale halves
04:11 PM anniepoo: 'odometry' strictly speaking is measurinng the robot's motion - eg with a sensor on a wheel
04:11 PM anniepoo: visual odometry is doing it with a succession of images.
04:12 PM SpeedEvil: Which also doesn't give you absolute unless you know what the wheel diameter is
04:12 PM anniepoo: of course
04:12 PM jandor: kool
04:12 PM jandor: I'll look at the sources
04:12 PM anniepoo: in many cases odometry reduces to SLAM -
04:13 PM anniepoo: simultaneous localization and mapping
04:13 PM anniepoo: Kozuch, I'd implement SLAM (using any of the packages available)
04:14 PM anniepoo: then take the image, segment the object you want to measure
04:14 PM anniepoo: and then project rays to locate the distance in the image,
04:14 PM Kozuch: well I think odometry is part (subset) of SLAM actually
04:14 PM anniepoo: and from that measure relative size
04:15 PM anniepoo: odometry data from a wheel sensor can be used to improve the pose estimation of slam
04:16 PM anniepoo: another good term - pose estimation - estimating the camera's position in a scene
04:16 PM anniepoo: slam is 'simultaneous' because it both localizes (solves the pose estimation problem) and maps (constructs a model of the scene)
04:17 PM anniepoo: having a good scene model helps with localization
04:17 PM anniepoo: good pose estimation assists in mapping
04:17 PM Kozuch: anniepoo, I do not need full SLAM, I am fine with the odometry part only
04:19 PM anniepoo: as I understand it, the two processes operate in feedback.
04:19 PM anniepoo: but if you just need pose estimation, there are non SLAM pose estimation algorithms
04:19 PM anniepoo: some of those don't identify objects in scene at all
04:20 PM anniepoo: they just identify features, then map the motion of features
04:20 PM anniepoo: eg if you track in with a camera, features move radially out from the aim point
04:21 PM Kozuch: anniepoo, yes I know this
04:21 PM Kozuch: but I dont know much about the scale problem...
04:21 PM anniepoo: so you have some object located in the image at time t
04:22 PM anniepoo: and the same object at t+1, and you know how the camera moved?
04:22 PM Kozuch: SpeedEvil, regarding the camera height (above ground?) - you say scale halves with double camera height. do you know the reasoning behind this?
04:23 PM SpeedEvil: Kozuch: the camera-ground axis is your only reference
04:23 PM Kozuch: SpeedEvil, but how does the odometry code know "where is ground"?
04:23 PM Kozuch: I may be flying maybe...
04:24 PM Kozuch: or be upside down
04:24 PM SpeedEvil: Kozuch: I was assuming you were meaning for a driving robot
04:25 PM Kozuch: SpeedEvil, yes I was, I am just trying to generalize (even though I do not really need it).
04:25 PM SpeedEvil: you can in principle make a 3d model from optical flow.
04:25 PM Kozuch: SpeedEvil, can you recommend particular odometry methods or code?
04:26 PM SpeedEvil: the scale problem is not the hard bit
04:26 PM anniepoo: Kozuch - look up orthorectification
04:28 PM anniepoo: you can do odometry on UAV by orthorectifying and piecing images, then computing vehicle position from that transforma and some height measure
04:34 PM Kozuch: SpeedEvil, yes, I know my robots speed so I could do absolute scale 3D from motion (optical flow). However any 3D reconversion is said to bring some error to the game. I actually would like to know my position to a "reference frame" (which say may be the 5th frame from 10 frame sequence for instance). I was wondering whether I could at least know relative distnace from odometry before I jump to 3D. If relative scale was possible I could measure
04:34 PM Kozuch: distance with wheel encoders before my reference frame (say in frames #1 and #3 from 10) and then infer the distance from #3 to #5 from the "realtive" scale maybe...
04:35 PM SpeedEvil: GPS, ...
04:35 PM SpeedEvil: and yes, you would ideally want to jam everything into a kalman filter and have the weights of it all set appropriately.
04:35 PM SpeedEvil: you might even include external knowledge like the typical width of a door
09:51 PM rue_house: hmm, well I have a robot arm to wire up
09:54 PM * rue_house remembers a time before computers
09:57 PM rue_house: fear is just pesemistic anticipation
10:17 PM Anniepoo_: go rue! wire that arm
10:36 PM rue_house: sleee