#robotics Logs

May 23 2015

02:00 Jak_o_Shadows yop
02:57 just_corey Rue, your never here...
03:06 rue_house hey
03:07 rue_house so, did you cut anything yet
03:07 just_corey lol, my daily flash player has frozen message
03:08 just_corey no not yet.
03:08 just_corey no hacksaw in the apartment
03:09 just_corey i still want to build a bolt hobber
03:13 just_corey guess id better kill youtube before it eats all my available bandwidth and gets me kicked off this server
03:39 rue_house sory, head burried in a code window
03:39 rue_house akchgh, its 1am
03:40 rue_bed ok if I fall asleep I wont wake up with everything numb
03:40 rue_bed the new code to make the servo arm echo the trainer arm sucks
03:40 rue_bed I dont know how, but threr is a huge pile of latency
03:41 rue_bed for the fiar I might have to stick with the one that has the hard coded numbers
03:42 rue_bed I was worried about adding buffers like that
03:42 rue_bed 4 floating point buffers on a 2.5Ghz AMD and you get about .25 sec latency
03:44 just_corey hmm
03:45 rue_bed maybe its all just a farse, maybe computers never really did get faster over the past 15 years
03:46 just_corey the whole universe is a hologram anyway
03:46 just_corey dont try to bed the spoon, thats impossable.. instead try to remeber there is no spoon
05:05 veverak rue_bed: explain me what you are trying to do
06:30 copycat1 hi, i was wondering how robots move
06:31 Jinxit do you mean "how do robots know where to move?"
06:31 copycat1 its usually done with a move call to the motion control right?
06:31 veverak yeo
06:31 copycat1 so motion planning is based on a 3d coordinate system
06:31 Jinxit not always
06:32 copycat1 is that the only model for motion planning?
06:32 Jinxit it's not necessary to plan for height when you can only move in the plane
06:32 copycat1 so 2d or 3d coordinate systems?
06:33 copycat1 are there any motion planning models that can deal with ambiguity?
06:34 Jinxit many
06:34 copycat1 within a coordinate system you must issue move calls with precision, so you must do some calculations
06:34 copycat1 where can i read up on these?
06:34 Jinxit if you have a robot arm for example, they usually have more than 3 dimensions to explore
06:34 veverak copycat1: 99% of robotics algorithms are made in a sense
06:35 veverak that they work with propabilities
06:35 veverak :)
06:35 veverak copycat1: udacity got quite nice robotics course
06:35 Jinxit copycat1: look into rapidly-exploring random trees
06:35 Jinxit simple algorithm, cool results
06:35 copycat1 ah, where can an amateur go to read up on these?
06:35 Jinxit scholar.google.com
06:36 Jinxit if you can deal with academic texts
06:36 Jinxit otherwise normal google and youtube, whichever you prefer
06:39 copycat1 is there a low cost 6 dof robotic arm that i purchase to play with?
06:40 SpeedEvil http://www.ebay.co.uk/itm//131476926907
06:42 SpeedEvil Simulated arms would seem to count
06:42 SpeedEvil 6 axis seems likely to be very expensive - that's a lot of axis, and (oversimplifying) each axis multiplies the cost as it has to deal with all the forces from all the downstream ones, multiplied by a lever arm, and still be as precise as the end of the arm can do
06:42 copycat1 normally, how would one program such a robot?
06:43 SpeedEvil A popular method is to cheat
06:43 copycat1 my experience is limited to 2d gantry robots
06:43 SpeedEvil simply use an operator to program it
06:43 SpeedEvil and then follow that path at speed
06:43 Jinxit again, RRTs can work
06:43 Jinxit if you want actual planning
06:43 copycat1 when you say an operator, you mean someone has programmed a teaching interface?
06:43 copycat1 and i will be using that?
06:44 SpeedEvil To do it 'right' you need to simulate the arm, and the workpiece, and the desired effects on the workpieve and avoid collisions
06:44 SpeedEvil copycat1: yes - someone controls every action of the robot - and all teh joints using some controller, to program in the appropriate motion.
06:45 SpeedEvil then the robot simply does this at 100* speed, taking 5s not several minutes to do it.
06:45 SpeedEvil But it can now do that motion all day
06:45 SpeedEvil So the desire to work out a 'proper' motion path by some computational method goes away.
06:45 SpeedEvil If you're doing one-offs, then this of course doesn't work.
06:46 copycat1 sorry, i don't really understand what you mean
06:46 copycat1 you mean you teach it a programmed path
06:46 copycat1 then it will repeat that path and over and over?
06:46 SpeedEvil yes.
06:46 Jinxit telling it exactly what to do
06:46 SpeedEvil For example, painting a complex part on an assembly line
06:46 copycat1 so i will be doing that through a programming interface that someone built?
06:46 SpeedEvil the individual motion may be really complex, and need a half an hour to teach it. using a hand controller.
06:47 copycat1 i don't want to do that, i want to experiment with how a robot moves by making that interface myself
06:47 Jinxit search for robot manipulation planning
06:47 copycat1 and i'm guessing that interface would involve lots of mathematics?
06:47 SpeedEvil The other way is you simulate the workpiece, you simulate the robot and use a toolpath generator to work out the toolpath
06:47 Jinxit if you want to do it optimally you definitely need math
06:48 SpeedEvil then you check for collisions automatically, and you change the toolpath so that it doesn't snap
06:48 SpeedEvil This is several orders of magnitude harder to program, but much more flexible
06:48 copycat1 "snap"?
06:48 SpeedEvil break
06:48 copycat1 you mean if you program it wrongly
06:48 SpeedEvil explode, burst into flames.
06:48 Jinxit ~kaboom~
06:49 SpeedEvil If for example, the 'elbow' of the robot hits the workpiece
06:49 SpeedEvil when you're working on it with a tool held in the 'hand'
06:49 copycat1 sorry, i'm still confused about what level we are talking about
06:50 copycat1 as in, developing the program for teaching the robot
06:50 copycat1 if you do it wrongly, you run the risk of "breaking" the robot?
06:50 SpeedEvil Or the workpiece
06:50 copycat1 have you guys done that before?
06:50 copycat1 developing the teaching interface, not breaking the robot
06:51 SpeedEvil It's a _huge_ project - if you're talking about automated toolpath generation for flexible robots.
06:51 copycat1 well, it doesn't have to be optimal, it doesn't have to be automated
06:51 SpeedEvil It is fairly tractable for 2.5d ones like 3d printers.
06:51 copycat1 i just want to try stuff on it
06:52 Jinxit my recommendation is to do it in a simulator first
06:52 SpeedEvil If you're not programming each individual move, then it's automated.
06:52 Jinxit before you splurge on hardware
06:52 SpeedEvil yeah
06:52 Jinxit and do it on a simple model, like a 3 joins 2d arm
06:53 eedEvil wishes in some ways he'd gotten the nice industrial robot that went sort-of-locally for 400 q
06:53 SpeedEvil It was at least as tall as me, and 5 axis.
06:53 copycat1 hehe
06:53 SpeedEvil But, it had clearly been sitting out in a field for three years.
06:53 copycat1 most robotic arms come with a teaching interface right?
06:54 copycat1 are there any where you have the freedom to program the robot at a lower level?
06:56 copycat1 am i asking the wrong questions?
06:56 Jinxit you're starting your project in the wrong end
06:57 SpeedEvil First work out specific tasks or types of tasks you want to do.
06:57 SpeedEvil Then work out required capabilities to do those tasks.
06:57 SpeedEvil Then work out how to do those in the most appropriate way.
06:57 copycat1 there isn't any specific task, i just would like to try programming a robotic arm at the motor level
06:57 copycat1 then i work up my way
06:57 SpeedEvil Do you have several $K free?
06:59 copycat1 so if i don't want the motion control software, it would cost me more?
06:59 SpeedEvil No, that's the raw cost of a cheap robot, without the control
07:00 SpeedEvil http://www.ebay.co.uk/itm/DIY-6-Axis-Servo-Control-Palletizing-Robot-Arm-Model-for-Arduino-UNO-MEGA2560-R3-/351392328809?pt=LH_DefaultDomain_3&hash=item51d09d6c69
07:00 SpeedEvil unless you consider something like this useful
07:01 SpeedEvil This looks ideal for making dancing robot videos on youtube.
07:01 Jinxit lol
07:01 Jinxit accurate description
07:02 SpeedEvil In principle, something of this class might be useful, with additional video feedback, but I would hesitate to give it anything more than a small paintbrush
07:03 copycat1 gripping an object with a robotic arm?
07:04 SpeedEvil it has no gripper
07:07 copycat1 let's say i was a misguided newbie who was interested in buying it
07:08 copycat1 what would you recommend instead to get started with?
07:09 veverak folks
07:09 veverak somebody, and I suspected somebody present here
07:10 veverak made paper, that when you extrude material on 3d printed
07:10 veverak and it's not heated properly
07:10 veverak from your nozzle, you will actually get material with bigger diameter that is your nozzle diameter
07:10 veverak ie, for 0.4 nozzle you will get 0.6 output extrusion
07:10 veverak any tips or links?
07:18 Jinxit copycat1: make a simulator, or use an existing one
07:19 Jinxit (but for planning you need to know how to simulate anyways)
07:20 copycat1 in a simulator, i have a robotic arm and i can start programming it?
07:20 Jinxit that's the point yes
07:21 copycat1 ok :) have you tried using one before?
07:21 Jinxit not for arms, but I've written sims for vehicles
07:25 copycat1 thank you jinxit and speedevil for entertaining my n00b questions
07:25 Jinxit np
07:30 zaxxon Jinxit: in what did you write it, out of curiosity?
07:31 Jinxit since 100% accuracy wasn't needed, unreal engine
07:31 Jinxit working on a quadrotor now, and the slow update frequency (60Hz) is finally starting to bite me
07:33 zaxxon robot logic in the blit call? :)
07:33 Jinxit for now, yes
07:34 zaxxon ditto here
07:34 Jinxit should probably move it to a separate thread but it's doing alright for now
07:34 zaxxon emulated an arm in python/opengl (well, drawing lines in a 3D space with opengl)
07:35 zaxxon but the tickrate was stable enough for the 50hz servo update rate
07:35 Jinxit yeah
07:35 Jinxit UAVs want something like 200Hz ideally
07:35 zaxxon broadcasted the position updates via igmp, raspi received them and i2c'd them to AVRs which did the pwm dance
07:36 zaxxon (all fun & games for personal interest, nothing serious of course)
07:36 Jinxit is the raspi stable enough?
07:37 Jinxit I figured OS scheduling would introduce too much latency at times
07:37 zaxxon for my purposes it seemed to be, it only needed to live 20 times/second
07:37 Jinxit ah
07:38 zaxxon never benchmarked that against it's own clock, designed/printed the arm and then got sucked into a project for work :(
07:43 veverak http://i.imgur.com/ct2uhgH.jpg
07:43 veverak SWEEEEET :D
07:44 zaxxon (dat battery)
07:47 Jinxit veverak: does it fly?
07:49 veverak I hope so
07:49 verak want such a t
10:37 rue_bed veverak, I have a arm done with pots, that is used to 1:1 control a robot arm
10:38 rue_bed I recently redid the code that shovels the positions between the two
10:38 Tom_itx wearable?
10:38 rue_bed its quite nice and generic, configurable, and somehow slow
10:38 rue_bed no
10:39 rue_bed there is no standardized word for the damn thing
10:41 Tom_itx can i mount another bootable drive?
10:42 Tom_itx without screwing it up
11:21 rue_house it dosn't care if a drive is bootable
22:06 rue_shop3 just_corey, awake?
22:14 Jak_o_Shadows am watching this guy https://www.youtube.com/watch?v=KGBWS-Y8goQ make a gear
22:14 Jak_o_Shadows I had no idea milling gears took so long
22:14 Jak_o_Shadows also, I keep hearing his phone go off
22:26 Tom_itx that's what cnc's are for
22:27 Jak_o_Shadows I am strongly getting that feeling.
22:28 Jak_o_Shadows Does a cnc leave a different type of grain structure to doing it with a lathe/magic gear cutter thing?
22:29 Tom_itx https://www.youtube.com/watch?v=ZhICrb0Tbn4
22:32 Tom_itx https://www.youtube.com/watch?v=w9x7BjDhxSs
22:39 Tom_itx https://www.youtube.com/watch?v=L7i_QDehseg
22:39 Tom_itx milling instead of grinding
22:44 rue_shop3 hahah
22:45 rue_shop3 today I was so confused at how I'd flared a pipe, I ended up flaring the new one on the lathe
22:45 rue_shop3 I USED THE FLARING TOOL before, just worked it out
22:47 rue_shop3 youtube seems to have retracted their bincoin miner, I can suddenly play videos on everything again