Inqling Junior - Robot Mapping, Vision, Autonomy

Posted by: @robotbuilder

I was wondering with pointing the robot base in a definite direction how accurately that could be done with an electronic compass?

You've reminded me of my unfinished topic https://forum.dronebotworkshop.com/sensors-modules/inertial-navigation-system-ins-gyros-accelerometers-magnetometers/#post-33818 and I've started adding my observations on the Magnetometer there.  I think the detail there will explain this following statement -- I'm cautiously optimistic that the magnetometer can be used as you suggest.  I'm hoping that the compass data in conjunction with accurate dead-reckoning data in conjunction with the point cloud data from the ToF can be used together to achieve some accurate mapping.  Until I start bringing all that data together, I can't even guess what will be the most dependable data and which will be the "reality checks".

Posted by: @robotbuilder

Before GPS people used to use a hand held compass on walking treks. Of course they also used distant landmarks where possible. 

Damn, that makes me feel old!  😆 ... I used to do triangulation with compass, maps and mountain peaks on hiking trips.  Real men don't use GPS... John Wayne or was it Clint Eastwood  😋 

Posted by: @robotbuilder

I also noticed more cheap robot vacuum cleaners coming out now using gyroscope navigation to improve their performance.  I wonder if that would help supplement encoder data.

Have you seen anything beyond the marketing hype on these robots.  I really think it is a buzz word that they're adding to say my vacuum cleaner is better than yours without any real merit.  Like the magnetometer, I think it might be used to give reality checks, but in the INS thread, I show how the integrated readings coming off the Gyro drift very badly.  I can't see how it can be used effectively to do navigation. 

Unlike using the Gyro for robot balancing, where you have the accelerometer to reset the drift every once in a while, there is no way to reset the drift in the X-Y plane (Z being up).  Maybe if we can trust the compass to reset it, but if we can trust the compass, why use the gyro. 

I'm all ears if someone has a concept.

VBR,

Inq

It looks like I can't improve on the above configuration.  It does appear that the angle of rotation is the culprit.  

To give a reality check, I did the calculation on the last video in the auditorium where the return is off by about 2 inches.  Doing the trigonometry... when I specified a 180 degree turn, I actually got a 180.19 degrees. 

Back on the spin table calibration test, I could not get consistency in the results... Doing the same test over and over, gave a variety of final angles being under or over by fractions of degrees.  The the digital steps of the stepper motors are exact and so the Math is accurate to seven decimal places, so the variability must be I'm getting down into things like play in the motor shafts.  The wheels, although not perfectly round, don't appear to be slipping on the shafts and should be compensated by the calibration steps performed above.

I guess I'm going to have to live with that for dead-reckoning ability and make sure sensors and ToF data can re-adjust the paths.  But it does go to show how difficult a problem it is to track around a building and return near where it started and have to perform the closure adjustments to ALL the data gather during the mapping.

More to come!

VBR,

Inq

@inq 

I guess I'm going to have to live with that for dead-reckoning ability and make sure sensors and ToF data can re-adjust the paths. But it does go to show how difficult a problem it is to track around a building and return near where it started and have to perform the closure adjustments to ALL the data gather during the mapping.

Have you looked at the math and algorithms yet for making maps and using maps from TOF data given approximate and later unknown poses?

I have the chips below which I have yet to experiment with. Although I can see uses for them I don't see any reason to use them for exact navigation or location identification as I am fixated on using a visual system and don't need to balance or adjust for a wobbly robot base.

https://www.societyofrobots.com/sensors_accelerometer.shtml

To enlarge an image, right click image and choose Open link in new window.

Posted by: @robotbuilder

Have you looked at the math and algorithms yet for making maps and using maps from TOF data given approximate and later unknown poses?

Not in the time frame of this project.  But it's something I've done in the past doing 3D transformations for 3D graphics, Ray tracing and Finite Element Method programs.  Its pretty strait forward stuff, so I haven't gone searching for my old programs yet.  For me... the big elephant in the room are these things you know off the top of your head like making a motor move, make it move where I want, make it accelerate so it doesn't stall, working with sensors, what are their limitations and extremes and variability.  Just doing simple electronics doesn't come natural to me.  It's been a while, but I still let the magic smoke out of parts on a regular basis.  These are what give me fits.

The Math will just pop up in a post that I'll probably gloss over unless someone take an interest in it.  It's not particularly much code.  I have a generalized matrix/vector manipulation library somewhere that probably has a hundred lines... after that, taking the 8x8 distances from the ToF, an X,Y,Z of the bot in some global coordinate system and an azimuth of the bot and a further angle of the head should be a one-liner to start spraying the point cloud.  Visualizing it using some JavaScript on an HTML canvas will take some time.  Doing front-end web development is not something I did professionally.  

Posted by: @robotbuilder

I have the chips below which I have yet to experiment with. Although I can see uses for them I don't see any reason to use them for exact navigation or location identification as I am fixated on using a visual system and don't need to balance or adjust for a wobbly robot base.

Yeah... I can't think of a good use for these sensors since you are doing a visual system.  I still have a third post to the Mag topic.  Most of the newer mag sensors also have the accelerometer built-in.  They need that to sense down so that it can do the 3D transformation for heading calculation since the Earth's magnetic field isn't always parallel to the earth surface, it gets messy so its a good thing you got both of those.

VBR,

Inq

Posted by: @davee

Mechanical stuff is definitely something everyone else does better than me, so this may be ignorance on my part, but doesn't this give you problems on your 180 degree turns at the end of the hall?

 

I can understand making subtle changes to the drive to the left and right wheels to correct for slight changes in diameter on a straight line, but if the wheels have an assymetry around their circumference, then during a 180 or similar turn, it seems like it will be a matter of chance which fractions of the wheel are involved with each part of the vehicle turn.

Hence, empirically looking for a 'constant' correction factor could be doomed?

In fairness, you are looking for an extremely accurate control in that turn, so I am amazed you get as close as you do!

 😉Maybe it's Heisenberg's Uncertainity Principle you should worry about, rather than the infamous cat?😉 

Best wishes, Dave 

Like I thought, having my dirty laundry out on the cloths line for all to see is nagging me.  Thanks for that @davee 😋 🤣 I've left slop in my design which is flowing down-stream into electronic and software design.  That simply is not acceptable!

I have to find a more hardened shaft to wheel interface.  Fortunately the 3D Printer world has lots of stuff to interface with Nema-17 stepper motors.  I'm surveying, but something like this might be the ticket.

https://www.amazon.com/Coupling-Connector-Accessory-Fittings-Connector-Silver/dp/B07PMMPSXW/

VBR,

Inq

Hi @inq,

   Hope you find a solution that works .. and don't fret about it ... you are making and testing these great machines ... I am only playing the armchair critic, and they are two a penny ... I haven't even had the courage to add a 3D printer to my toy set yet!

Your solution reminded me of a small Meccano set I had many decades ago! I seem to remember it had part like that .. albeit maybe with larger flange .. that you could bolt rectangular panels onto, to make a windmill. Sometimes, the old solutions are still the best? Perhaps I should consider a new Meccano set before straying into 3D printer land? My mechanical prowess might just about stretch to using a nut and bolt.

Plus this might be a first for me ... from time to time I have found someone trying to fix a 'fundamental' electronic hardware flaw (due to 'sloppy' design) with a software 'correction' ... but this might be the first time for someone with mechanical 'slop' trying to corrrect it with software! 😉

Best wishes my friend .. take care but have lots of fun,  Dave

Posted by: @davee

Plus this might be a first for me ... from time to time I have found someone trying to fix a 'fundamental' electronic hardware flaw (due to 'sloppy' design) with a software 'correction' ... but this might be the first time for someone with mechanical 'slop' trying to corrrect it with software!

I'm not used to someone beating me to the mechanical engineering flaws.  Keep me on my toes or hold them to the fire when you can!  Thanks @davee.

VBR,

Inq

Posted by: @inq

3D Printers are the best toy ever... including sliced bread.  I regret not jumping in earlier.  The idea that I can draw something and press a button and start using it, often in the same day is just magical.

I agree totally, plus I'm lousy with hand tools, so having something that makes things the right size and puts the holes exactly where they're supposed to be is an incredible boost !

BUT - If you're not mechanically inclined, I'd caution you against them.  They are not plug-n-play like an Inkjet printer.  There is a learning curve with hardware and software.  And they mess up and have to be fixed.  I've had to repair mine a couple of times and its a premium printer.  I've never heard of a repair store for them, so you have to take them apart, tune and fix them yourself.  Parts are dirt cheap, mostly interchangeable, but you have to figure which ones you need and how to fix it.  I'd imagine there is plenty of guidance on the forum, but you still have to use screwdrivers and wrenches.

I agree with this even more, I've had 9 printers over the last 10 years, 3 have died from obsolescence and/or overwork, one's on its death bed and one of the remaining ones is at the borderline of requiring wok. Working on them never really ends, but the star of the lot has had well over 150 pounds of plastic shoved through it.

For me, it's a love-hate relationship, I hate tinkering 🙂

@will Name that star PLEASE!

@zander 

Cetus Mk 1

Note that this is an answer to a query, NOT an endorsement !

@will Noted.

@zander 

I suppose I should explain that ...

1) the Mk1 model is no longer available

2) Cetus is now charging $60 for a single nozzle. Since nozzles wear down and are therefore considered consumables, I have spent more money on nozzles for this printer than I paid for the printer. This offends me greatly (I'm a cheap bastard) so I won't recommend Cetus printers (nor will I buy Cetus printers to replacements the three Cetus ones I already own),

I've only had two.  The first was POS!  Total ignorance on my part and not knowing who to turn to for advice.  The second has been my work horse.  I'm too lazy to calculate the weight and number of rolls even is fleeting.  I can't remember what I had for breakfast.  But mine has:

@inq 

Looks like you got your money's worth out of it 🙂

Posted by: @will

2) Cetus is now charging $60

Damn... Did you even get kissed for that?

I get the dirt cheap ones for mine.  Never had a lick of trouble with nozzles.  Other things, yes.

https://www.amazon.com/Printer-Nozzles-Hardened-Extruder-Cleaning/dp/B09537J84X/

VBR,

Inq