#robotics Logs

Apr 02 2015

06:13 orlock char inChar = (char)Serial.read();
06:13 orlock if (inChar == ACK)
06:24 deshipu orlock: that question is dumb allright, I can't figure out what it means
06:25 orlock yeah i didnt finish it
06:25 orlock got it sorted out
06:25 deshipu a cliffhanger
06:25 deshipu don't do that :(
06:28 malinus haha
06:29 orlock The question was - This code doesnt work, what should i have in place of ACK for that ascii value
06:29 orlock the answer is 6
06:29 orlock for the raw numeric value of ACK
06:29 malinus :)
07:16 Tom_itx or #define ACK 6
07:16 Tom_itx then use it
07:18 malinus enum{ACK=6, CR=13}
07:18 malinus :)
09:42 blib malinus: what are you upto these days?
09:44 malinus blib: trying to figure out how to make some mechanical stuff for my robot. Never really made anything mechanical before. What about you?
11:45 blib malinus: trying to put some sensors on my robot :)
11:45 blib need to find what to buy so that later on they don't give me pain
11:50 deshipu blib: what do you want the robot to do?
12:59 blib deshipu: just not run blindly and be able to process the sensor information to do some cool tricks?
13:00 deshipu blib: I suggest you do some reasearch and plan some more details...
13:00 deshipu blib: knowing what you don't want to do is useful, but insufficient
13:01 deshipu brb
13:30 blib deshipu: I want to avoid collisions. Second thing I want to do is to find out the directions (N S E W), and want to know how fast I am moving/accelerating and in which direction.
13:31 blib deshipu: I am hoping to get some help here, before I buy parts etc.
14:03 deshipu blib: what kind of obstacles?
14:03 deshipu blib: you want to do SLAM?
14:05 Loshki blib: budget?
14:12 blib deshipu: which SLAM code is useful on ubuntu?
14:12 blib Loshki: cheap (Less than 1k) - and I want to run it outside
14:14 Loshki blib: check out Neatos, and robomower technology...
14:16 blib Loshki: I am using a raspberry pi. So need everything running on ubuntu + pi.
14:16 blib Loshki: how much do I have to spend on a decent lidar sensor for collision detection, if I don't rip it out of a neato?
14:18 Loshki blib: I see neato lidars on ebay for about $70. For $100, you can get a working Neato to hack on. Not outdoor-proof, but still, 80% of what you need already debugged and working. And if you're lucky, it vacuums too...
14:19 Loshki malinus: I found Cooks "Intermediate Robotics" jammed with useful construction info.
14:20 blib Loshki: if I want to buy an outdoor lidar, how much and which one?
14:24 Loshki blib: The lidar does work outside, though full sunshine may dazzle it, but it's not weatherproof. I don't know of any lidars priced for hobbyists that are guaranteed to work outdoors. The design has moving parts and is open to space, as it were. Ultrasonic sensors instead? -- Very much the poor man's version...
14:33 blib Loshki: if money was not a concern, what would you use?
14:34 blib also, I've a small 6000mAh battery that I want to last. So power is definitely a concern
14:35 Loshki blib: it's not a fair question. Money is always a concern. What exactly is your application?
14:42 Loshki blib: Also "I am using a raspberry pi". Way too soon to be deciding on hardware. Work on your requirements first.
15:03 malinus Loshki: do people never use lasers for range detection?
15:03 malinus (in robotics)
15:03 malinus I guess it's kinda cumbersome
15:04 malinus oh nvm
15:05 malinus I thought we were still talking about tof for some reason.
15:06 deshipu you know what would be more cumbersome?
15:07 deshipu shooting raspberry pis at the obstacles and trying to determine how far they are by how they bounce
15:07 Loshki malinus: Neato lidar isn't true TOF. It actually uses triangulation...
15:07 deshipu you could also just put a Kinect on it
15:11 chris_99 if you want to measure to 25cm you can get an ST chip :) which i've played with
15:11 chris_99 and is proper ToF
15:11 chris_99 using an ir laser
15:12 malinus chris_99: haha I thought it was because they were so light and cheap. So throwing them wouldn't be that expensive.
15:13 chris_99 vl6180 is the one i mean
15:13 malinus yeah, someone posted a video with a presentation of one of those
15:15 malinus Loshki: I don't really need to help building the robot itself as such. It's more the mechanical part of a contraption that I want the robot to have. I kinda want to make a treat launcher :P.
15:15 SquirrelCZECH deshipu: that raspberry pi distance-meter
15:15 SquirrelCZECH makes me feel like better urasge of raspbi than most of the things I saw
15:16 malinus heh
15:20 deshipu SquirrelCZECH: well, we don't judge here ;)
15:20 Loshki malinus: check out treat launchers on instructables/hackaday etc. Rape and pillage (of ideas) is your friend...
15:21 SquirrelCZECH deshipu: O:)
15:23 deshipu I found that there is a lot of poor literature on robotics, but quite few books that are actually useful, but still directed at newbies
15:24 deshipu I mean, too few
15:24 deshipu either you are studying engineering, reading published papers, or gtfo :(
15:26 deshipu and this stuff isn't actually that hard
15:26 Loshki I find the same problem with robotics as with most other subjects. Tons of stuff for beginners, tons of research papers for academics, and virtually nothing aimed at what I consider the 'intermediate' level.
15:27 deshipu there is http://www.dspguide.com/pdfbook.htm
15:27 deshipu which has a lot of useful stuff for sensors
15:28 deshipu and interpreting the sensor data
15:41 adam789654123 wow!!! looks like a great book!
15:47 deshipu https://www.youtube.com/watch?v=ENHvCGrnr2g
15:47 deshipu nice walking!
15:48 adam789654123 makes sense, the way it balances
15:48 adam789654123 on three legs while planting the other
15:50 adam789654123 looks like they used the motion of four legged animals for inspiration
15:51 deshipu adam789654123: creep and trot gaits
15:51 deshipu adam789654123: those are the two basic ones
15:51 adam789654123 well.... the different gaites
15:55 blib Loshki: I am using rpi because I'm familiar with ubuntu, which I can run. My application is to make an autonomous robot that can roam around outside.
15:56 adam789654123 when four legged animals run though.... their front and pack pairs of legs work in unison
15:56 adam789654123 when four legged animals run though.... their front and back* pairs of legs work in unison
15:56 adam789654123 or
15:57 adam789654123 when four legged animals run though.... their front pair works as if one leg, and their back pair works as if one leg
15:57 adam789654123 they syncronize
16:00 deshipu adam789654123: that's gallop
16:01 deshipu adam789654123: it's pretty different from walking
16:01 deshipu easier in some regards
16:07 adam789654123 well..... now i know the difference between a gallop and a trot
16:09 deshipu the problem with running is that you need quite some strength and speed in your actuators for it
16:09 deshipu creeping can be done at any speed, as it is statically stable
16:09 deshipu you can even stop at any point
16:10 deshipu and it's not like you can't do a lot of stuff with creep
16:10 deshipu like that step reflex, for a start
16:17 adam789654123 yeah.... and running efficiently requires alot of spinal motion
16:20 deshipu not inherently
16:21 deshipu you can run on one leg with two degrees of freedom
16:21 deshipu spinal motion lets you do it with less energy
16:26 adam789654123 maybe i should have said, "running more efficiently ..."
16:29 adam789654123 well.... anyway, the spinal motion of a running animal is impressive
16:30 adam789654123 i think thats the point of what im trying to express
16:30 deshipu yeah
16:34 malinus I wish I knew about openSCAD before
16:34 deshipu openscad is great
16:35 malinus yeah, it's so easy to put stuff together.
16:35 deshipu I need to try some blender plugins
16:36 deshipu that try to be openscad-like
16:36 deshipu because openscad's rendering is poor
16:36 malinus deshipu: do you use it for general cad'ing too. Or just smaller stuff/parts?
16:37 malinus deshipu: also btw. you can import openscad to freecad
16:37 deshipu malinus: cool, I have to try that too
16:37 deshipu malinus: I mostly use that for generating images for the instruction manual
16:38 deshipu malinus: like here: http://tote.readthedocs.org/en/latest/gait.html
16:39 malinus deshipu: wow that's sexy!
16:39 malinus deshipu: you made that servo yourself?
16:40 malinus deshipu: didn't this take ages: http://tote.readthedocs.org/en/latest/_images/dice.png
16:41 deshipu malinus: https://bitbucket.org/thesheep/scad-parts/src
16:41 deshipu malinus: not really
16:41 deshipu malinus: note how many things repeat in there
16:42 malinus yeah, and when you have the module, you just move it
16:42 deshipu plus, you can refine them and add details later
17:03 blib sorry, got disconncted. Is thre a TOF for more than 25cms (works day+night outside) that is < $500 and works outside?
17:04 ben_m Oh sweet, this channel exists :D
17:04 ben_m Hello!
17:06 ben_m Any of you guys have recommendations to wirelessly detect direction from one object (robot) to another (remote control)?
17:13 Hyratel ben_m, a bunch of small directional antennas
17:13 Hyratel or an RDF loop
17:13 SquirrelCZECH nah
17:13 uirrelCZECH need wireless charging for his la
17:14 Hyratel however an RDF loop is two possible direction results
17:14 Hyratel NE-SW, E-W
17:16 ben_m I'll look into these things, kind of a noob at all of this :) Thanks.
17:16 ben_m I tried it using Bluetooth (RSSI/signal answer dt) but that's just way too inconsistent.
17:36 malinus ben_m: what about sending gps data between each other?
17:36 malinus :D
17:36 ben_m How accurate is that?
17:37 malinus well, like idk. 5-10m with good coverage?
17:38 malinus I'm pretty sure it's a terrible idea though!
17:38 ben_m It's an idea at least!
17:38 malinus I guess it could work outside, on big fields.
17:38 ben_m Yeah it's all outside.
17:39 ben_m I'll be back in here soon to ask about affordable depth information cameras that work outside :^)
17:42 ben_m How would you use a 'bunch' of small directional antennas? Triliteration based on when you get a signal back?
20:01 armyofevilrobots Woot. X-axis moving on new printer :D
20:01 armyofevilrobots Difficulty: from scratch new printer design, including motion control board and firmware.
20:08 Triffid_Hunter armyofevilrobots: I designed my own electronics and firmware when I made my first printer
20:08 Triffid_Hunter and I'm currently designing a delta
20:10 Triffid_Hunter armyofevilrobots: https://picasaweb.google.com/lh/photo/hq6v_oZ5fo3z6l42PHAAPtMTjNZETYmyPJy0liipFm0 and https://picasaweb.google.com/lh/photo/oxSHmuY2Y84gtsdi2Tp5dtMTjNZETYmyPJy0liipFm0 and http://github.com/Traumflug/Teacup_Firmware if you're curious
20:13 armyofevilrobots Delta is next, and yes, thanks!
20:14 significance Hello all! Is anyone here familiar with Kalman filters/
20:14 significance ?
20:14 Triffid_Hunter armyofevilrobots: I don't develop teacup anymore, I've moved on to other projects, but Traumflug is doing a great job as primary developer
20:14 Triffid_Hunter significance: I haven't actually played with them, but I've read about them heaps
20:16 significance Thanks Triffid_Hunter - I'm playing with extended Kalman filters, and I'm trying to get circle tracking set up. What might my state variables be? I was thinking (xc, yc, r, theta, v_theta) where (xc, yc) is the center of the circle, r is the radius, theta is current angle, and v_theta is angular velocity. Does that seem like a good start?
20:19 Triffid_Hunter significance: hm, polar kalman sounds interesting, might need a term for radial velocity too (towards/away from center)
20:19 significance Triffid_Hunter: good call. My big question is: if I have the update equations hold xc, yc, and r constant, how do I ensure that the Kalman filter figures them out properly?
20:19 Triffid_Hunter significance: what's the application? you could expect everything to explode if you move in a straight line
20:19 Triffid_Hunter significance: if they're held constant, they can't be changed and thus the filter can't "work them out"
20:20 Triffid_Hunter significance: for most applications I'd suggest just sticking with linear state vectors and calculating polar if necessary from lateral acceleration
20:21 significance Triffid_Hunter: In my transition matrix, the xc and yc shouldn't change right? But if they don't change, will the Kalman filter figure their values out by choosing values that minimize error?
20:22 significance Triffid_Hunter: I'm doing a computer simulation (last optional challenge for online course) so I sadly can't add sensors.
20:23 Triffid_Hunter significance: shouldn't be too hard to simulate some sensors with gaussian noise and see what shakes out
20:25 significance Triffid_Hunter: I've tried adding an acceleration term to a linear KF - is that what you mean by lateral acceleration to calculate polar? Sadly, the acceleration keeps changing to point inward, so it does a poor job of approximating...
20:26 Triffid_Hunter significance: of course the acceleration points inward for a circular path, that's how it works ;) should be able to do math with forward velocity and inward acceleration to find radius
20:30 Triffid_Hunter I bbl
20:30 significance Triffid_Hunter: Ah, that's smart. So I'll do a radius calculation using centripetal acceleration and use that radius to get the center of the circle from the localized robot (x, y). Really dumb question, but from here how do I localize my robot while keeping to linear matricies?
22:54 solifugus Any recommendations on a pneumatic manifest for humanoid?
22:55 rue_house good luck ont he valves
22:55 rue_house unless you have $250k
23:05 solifugus maybe one of these: http://www.smcusa.com/top-navigation/products/directional-control-valves.aspx
23:06 rue_house if you have $250000 or more, sure
23:08 solifugus they mention automotive applications.. i doubt it'd be that much, as a part for a car
23:12 rue_house heh
23:12 rue_house you know you need a hundred or so valves for a humanoid?
23:13 solifugus 32 will work.. plus a handful of steppers..
23:13 rue_house 32 actuators is usually 64 to 128 valves
23:18 solifugus I requested a ball park price range figure.... we'll see what they say. If its reasonable, great.. if not, I'll move on for an alternative.
23:19 solifugus it's a simple humanoid.. flywheel for stability, allowing simple actuations for locomotion.
23:20 solifugus raspberry pi is enough cpu..no actual balancing neural net required.
23:22 solifugus all rigid pvc skeleton.. heated pvc resin.. molded..
23:23 solifugus strong, light, cheap, non-abrasive... so no bearings needed on joints.
23:24 solifugus more robust than aluminim... almost like brass, in some respects.
23:25 solifugus melts nicely for molding at 90 celsius
23:26 rue_house mine is called Sparrow1 and its 12' tall, uses air muscles
23:26 rue_house its as small as I can make it from what I can get
23:26 rue_house budget is $240
23:26 rue_house currently scavanging parts
23:27 rue_house looks like I'm gonna put togethor the upper torso first, cause I need to put it to work
23:27 solifugus using individual solenoids?
23:27 rue_house current design challange is cheap pressure sensors
23:27 rue_house 188 washing machine valves
23:27 rue_house ^^ 'currently scavanging parts'
23:28 solifugus could use cheap rheostats on joints, implying pressure.. sort of.
23:28 rue_house I have a few ideas I'm playing with
23:28 rue_house did you see the disney actuators?
23:29 solifugus nope.. enlighten me..
23:30 rue_house http://hackaday.com/2014/10/10/ask-hackaday-who-is-going-to-build-this-pneumatic-transmission-thing/
23:33 solifugus neet..
23:57 rue_shop3 linear differential variable capacitor