Sunday, February 17, 2013

153> Railbotics!



Back at Christmas, I posted about having purchased a hub motor in the form of a cheap ceiling fan.  That acquisition marked a first tentative step toward building an experimental scale model of the SMART (Suspended Multi-axis Automated Rail Transport) PRT platform.  I thought an update might be in order.  After great deliberation, I decided that the fan motor would be just too hard to modify in the way that I wanted.  It was a little on the big side anyway. (1/4 scale) What I opted for instead was to completely start from scratch.  After all, what I really want is an ironless axial flux hub motor and they don’t make such a beast.  Sure, there are a few small “hub motors” out there, but they are generally of the radial flux design.  (Requires too wide of a wheel) And these days any motor that fits into a wheel’s hub is called a “hub motor” even though the motor might be stationary instead of turning with the wheel.  Some even have internal gearing, so they are not even one-piece devices.  These variations will only confuse everything and everybody.  In the end it turns out that the same constraints drive both the actual design and the scaled down version.  A taller, skinnier and shorter bogie simply allows a more compact track.  It seemed better to clarify these relationships with a model that is as faithful to the full scale version as possible.  Here is what I have so far, mounted on a demonstration stand.  I’ll dissect it for you in a moment, but first let me tell you one more reason why I am going to all of the trouble.

Ever see the video of that old tape by Aerospace Corp?  That is more than a good primer on PRT. That model is one very cool toy…There are still thousands of model railroad enthusiasts out there who get great satisfaction building similar stuff, not to mention robot buffs.  Could model “railbots” (not to be confused with the game) ever gain a following?  After all, a working model of the SMART PRT platform would certainly be equally fun to watch and vehicle avoidance and routing strategies would offer programmers something truly challenging to chew on.

Every year they hold competitions to race ridiculously impractical and expensive solar vehicles across the desert - the only place where they reliably work - even in the daytime - while teams of enthusiastic high-school kids build robots that can only push balls around.  What if these people had something a bit more practical to work on?  Beyond that, what could be cooler than a “train set” where the “cars” can go in any direction, even straight up?  And talk about educational!  This cross between model railroading and robotics would be wonderful for our high schools and colleges.  How about a race where the object is to carry full glasses of water through a 3D network without spilling a drop or touching another vehicle?

In the course of looking for components, I ran into all kinds of hobbyist robot motors – but no gearless, direct-drive systems.  Maybe there’s a market for such things… perhaps enough to where I could get some much needed help refining the control electronics and software, or to where someone might even start offering something like this in kit form, or even a whole set.  I can see it now - Lego Mindstorm's new “Mobot Racer” kits. Every kid should have one!

I have seen model trains chugging around restaurants and malls.  I’d love to donate a “railbot” set for the cafeteria at Google.  That would get them thinking!  I’ll bet they would be delivering food to the tables in no time.  Finally I would add that the I have spent a lot of time considering ways to make SMART PRT track affordable, and these attributes play out in the planned 1/6th scale as well.  Making homemade track should prove both easy and very cheap. But making these little guys fast as slot cars yet able to climb straight up without a bogey body fat with gears takes some slick conversion of electromagnetic force.  Hence my little project.  


Here is all you need, propulsion-wise.  This is all made with hand tools and stuff like lamp parts, PVC pipe caps … nothing more exotic than motor wire and magnets.  The shaft and winding assembly (left) is simply sandwiched between the two magnet arrays (bottom) and capped with the piece shown on top.

Below is the heart of the motor.  On the right are bare windings and on the left I am holding similar windings cast in a polyester resin disc.  I embedded aluminum window screen on both sides, both for strength and for a heat-sinking purposes.  In larger scale production the windings would be captive in an aluminum disc designed for the purpose.  I produced the windings by taking 24’ (4.2 ohm) lengths of wire and winding them around a spool (washer/square nut/washer arrangement) spun by a drill… about 6 minutes each.

The basic principle is this.  There are 16 windings sandwiched between 20 rotating magnet sets, which creates 64 positions per revolution where 4 magnets and 4 electromagnets can align. (They form a cross pattern).  The individual windings are series connected in four cross-shaped groups. For example, windings at the north, east, south and west positions are connected together with a single set of leads. The magnets alternate in polarity and so do the individual winding leads.  Sorry – some people will actually WANT these details…

The bottom line is that each winding set can be electrified to attract nearby magnets and when they reach dead center the winding reverses and repels that magnet instead.  Meanwhile other magnets are at play, attracting or repelling hard because they are all close to, but not on, their own dead center positions.  Each group of coils reverses polarity 16 times per revolution.  To run the motor (in the most obvious fashion) requires 4 separate, consecutively timed square waves, so that makes 64 total polarity changes per revolution.  There are probably all kinds of ways to tweak those waveforms for more speed with less heat… The way I outlined is “full on” for maximum torque.  Two sequenced bipolar stepper motor drivers should run this nicely and enable micro-stepping, which would both smooth out the wave form and allow the coils a bit of lower voltage time to cool.  Not that I would use necessarily use this like a stepper motor, although it certainly can be. Since the magnets are of alternating polarity, a Hall effect sensor place almost anywhere will trigger 20 times per revolution, so servomotor style feedback will be accurate to that (18 degree) fraction of a turn as well, as is.  Between these two forms of position determination and markers in the track, our little bogie will never be lost – even by fraction of an inch.

Notice the little blue circuit board in the picture below.  That is an Arduino microprocessor board, which I will use to supply the root clock signal and direction information to four such wheels.  It will also take sensory data and branch the programmed motor instructions appropriately.  Also note that I have added a sprocket for cogged (no-slip) climbing.   Finally, since the case is made from standard 4” drain pipe caps, its perimeter can be conveniently covered by a slice of rubber sleeve (sold as a flexible pipe connector) so my little wheels will even have durable rubber tread.




Now for the hard part.  The analog electronics and software programming.  Unfortunately I’m not trained in either, or this little puppy would be going through its paces right now.  Lately I’m up to my ears in Googled articles about “H bridges” and transistor saturation.  At least I have plenty of holding torque and I’m continuing my education!