Modular Drive System

[DavisDad]

After two FRC seasons, I've been thinking about a design for the drive system with the following characteristics:


Modular- self contained unit that could be used year-to-year regardless of the chassis design
Compact- our previous designs have taken up a lot of "real estate" in the chassis
Reliable- both mechanically and controls
Controllable- teleop control smooth and predictable
Agile- not jerky
Good efficiency
Strong- takes a lickin' and keeps on tickin'
Low maintenance


My son and I have been working on an FTC system that is performing well:

http://simhardware.org//img/BeaverBot_small.JPG


We built the prototype with a direct drive from the motor for testing. The FTC components have:

http://simhardware.org//img/DrivePrototype.JPG


MATRIX motor with planetary gearhead and encoder
MATRIX motor controller uses encoder for PID speed control
4" VEX Mecanum wheels
Omni-directional controls programming for Mecanum wheels


We're working on an FRC version:

http://simhardware.org//img/FTC_DriveAssembly.JPG

Full size CIM motor
Encoder
Jaguar CAN bus speed control
Wheel hub planetary reducer, 5:1 (think Sturmey Archer bike hub)
Bevel gear reducer 2.5:1


I'll post our progress

[CENTURION]

Looks like a good start!

But, please keep in mind that unlike FTC, FRC has rules against reusing systems from year-to-year.

[DavisDad]

Hi CENTURION,

Thanks for the heads-up. I'll look into that. I'd assume FRC would encourage teams handing down engineering design and build information year to year. I'll get up to speed on the spirit of the rule.

Much appreciated,

Craig

[MrBasse]

As long as you make your design public (which you just did) and you make new ones during the build season, you are generally okay. Just don't use anything that has been prefabricated. Of course this years rules could always be different...

[Woolly]

As long as you make your design public (which you just did) and you make new ones during the build season, you are generally okay. Just don't use anything that has been prefabricated. Of course this years rules could always be different...

Well, if you were to have a company that made modules like that and sold them pre-assembled, you could classify it as a COTS item and re-use it.

[MrBasse]

Well, if you were to have a company that made modules like that and sold them pre-assembled, you could classify it as a COTS item and re-use it.

To do that it has to be available to everyone, everywhere. That is a heck of an undertaking if they want it to stay the way it is as they designed it.

On the upside, this would have to be rebuilt anyway to make it suitable for FRC use, right?

[Woolly]

To do that it has to be available to everyone, everywhere. That is a heck of an undertaking if they want it to stay the way it is as they designed it.

On the upside, this would have to be rebuilt anyway to make it suitable for FRC use, right?

Absolutely!
Though, this may just be me, but if I'm going to build modular drive units, and I have experience with CAN bus encoder feedback control, I would build swerve modules that aren't specific to which corner of the robot it's placed on, such that I could build 6 modules per robot I build and have spares that can be placed on in the event of a module failure.
Otherwise, I would look at maybe an octocanum system that uses pneumatics to switch between the mecanum wheel and a traction wheel.
Experience tells me that Mecanum doesn't exactly help you in the eliminations draft, as teams get wary of defense being played on you or you being unable to play defense.

[DavisDad]

Absolutely!
Though, this may just be me, but if I'm going to build modular drive units, and I have experience with CAN bus encoder feedback control, I would build swerve modules ...

I'd love to build a swerve drive. I agree it's the best technology. But, we can't afford a Cadillac. :)

The prototype we've built works reasonably well given the relative simplicity vs. the swerve design. We tested the prototype with 4" regular wheels and the reduction in power (tank mode, forward) wasn't too bad. Our thinking is to make the drive system very doable for the team and allow more time/resources for the game functions.

I'm bugging my son to make a video of the Mecanum teleop to post. The kids are able to control the thing with amazing agility. I guess the years of video gamming is good for something. ;)

[DavisDad]

"One of the purposes of the FRC is to provide Team members with the experience of conceiving, designing, and constructing their solution to the annual competition challenge. We want each student to have the experience of creating a new system each year... Solutions that merely bolt together a minimum number of externally-designed COTS subsystems may not offer the students the opportunity to understand the “why” or “how” of an item’s design. Likewise, solutions that are merely minor modifications of a design utilized for a previous competition does not offer the current students complete insight into the full design process"

Re-reading the FRC manual from last year, what do y'all think of this approach:


Maintain good design documentation; drawings,, specs, fabrication details...
Review previous design vs. current game requirements: gear ratio, speed vs. power, agility, etc...
Keep previous year's design elements if no advantage in changing system


I'm sure each team will add improvements if basic design is retained. I don't see any difference in using AM Mecanums & ToughBoxes from previous year's design. There's plenty of opportunity for the students to experience the engineering process with the other game functions. After years of building clean-rooms, I avoid "reinventing the wheel" whenever possible. ;)

I guess the key is:

"Purchasing optimization and design re-use are both important concepts; however, Teams must be cautious not to over-utilize them to the point that the student’s experience is compromised."

[CENTURION]

Yup, probably one of the hardest things for me when I was starting into mentoring is to remember that, unlike in industry, the point isn't always to get the thing done the fastest and most efficient way. The point is to teach why it's done that way, even if it means doing it less efficient way, and doing it that way over and over.

I usually make the student do it the hard way the first time (or the first couple times). And then, once they've done that, show them the easy way.

I make students tap things freehand when they're new, and then once they've tapped four or five parts like that, and they get frustrated from having to constantly make sure that the tap is going in straight, I introduce them to the tap guide block that does that for you. ;)

[Joe Ross]

As long as you make your design public (which you just did)...

I don't think that posting a couple of screenshots counts as making the design public as far as reuse goes. In my mind, a cad model or a set of dimensioned drawings would be required.

Unfortunately, last years rule requires that the design be public before kickoff, even though the rules and any chance for clarification aren't available until after kickoff. I think that any rule that governs something before kickoff should be available before kickoff.

[MrBasse]

I don't think that posting a couple of screenshots counts as making the design public as far as reuse goes. In my mind, a cad model or a set of dimensioned drawings would be required.

Unfortunately, last years rule requires that the design be public before kickoff, even though the rules and any chance for clarification aren't available until after kickoff. I think that any rule that governs something before kickoff should be available before kickoff.

I agree about the after kickoff rule issue completely. I looked at their images and I'm pretty sure that I can build that without more information, which speaks to the simplicity of their design.

Either way I don't think that part would matter, as these modules as pictured wouldn't be up to the task of FRC in my opinion.

[DavisDad]

I agree about the after kickoff rule issue completely. I looked at their images and I'm pretty sure that I can build that without more information, which speaks to the simplicity of their design.

Either way I don't think that part would matter, as these modules as pictured wouldn't be up to the task of FRC in my opinion.

We'll be publishing all specs, drawings & code as design progresses. If anyone would like specific information, let me know...

Were not posting design to meet FRC rule requirements; intent is to share ideas for collaboration. There are many pesky details, particularly with coding, where solving problems could benefit everyone. We've yet to work out the CAN Bus/Jaguar CAN Bus encoder built in control functionality to do what we've done with the FTC NXT/Matrix prototype.

[DavisDad]

Yup, probably one of the hardest things for me when I was starting into mentoring is to remember that, unlike in industry, the point isn't always to get the thing done the fastest and most efficient way...)

CENTURION,


Excellent advice! It's a beautiful thing to see the kids go from almost no fabrication ability to understanding, and then ownership of the build. It's very rewarding when they "push me out of the way" and take off on their on.

[DavisDad]

...I usually make the student do it the hard way the first time (or the first couple times). And then, once they've done that, show them the easy way...

This reminds me of a time, right out of school, when I was doing some simple machine design work. I would bring my half-baked shop drawings to an "old school" machinist who was trained shortly after WWII. He was very kind to me and spent time going over the drawing details and explaining why my tolerancing was poor. He'd make one beautiful hand drawing in seconds that contained more information than my 3 CAD drawings I'd spent hours on.

One day when he'd accepted my drawing with no corrections, he said:

"Craig- let me show you something"

He brought me into the back of the shop, pulled back a curtain revealing a part he was working on; an amazing steel object that looked more like art than machine. He said:

"My first task in machinist school was to take a file, vice and piece of steel and make a machine screw to spec tolerances. Now at the end of my career, I'm given a high-tech, super-computer design that has no geometries that can be mathematically derived. I'm roughing the shapes with the CNC and am finishing by hand filing. The young machinists want to skip the manual work and do everything by computer..."

This was one of the most enlightening experiences of my life.

[DavisDad]

...as these modules as pictured wouldn't be up to the task of FRC in my opinion.

MrBasse,

Could you say what about the design you think is weak?

Thanks

[MrBasse]

MrBasse,

Could you say what about the design you think is weak?

Thanks

I don't think the design itself is weak, simply that the current version posted would be up for quite a task. I have no clue as to speed, but I would be concerned with the gearboxes (we have found planetariums to be a tad fragile) in a pushing match. An open frame with little structure to tie the front end together would have me concerned as well for impact, but I think with the scale of your frame rigidity shouldn't be much of an issue. Again, these are simply in the prototype you have proposed. I haven't a clue what this could delvelop into when built for FRC. I am excited to see what it could become, as a modular approach is something we have been contemplating for our entire history in FRC.

Beyond that, it's a tiny little thing! Sure does make the battery look massive! (I know it is a prototype!)

[DavisDad]

I guess my post is unclear. The prototype is the FTC team. We're scaling up for FRC. The FRC battery was used to add weight to simulate future game components. The open frame was min structure for testing drive performance

[MrBasse]

I guess my post is unclear. The prototype is the FTC team. We're scaling up for FRC. The FRC battery was used to add weight to simulate future game components. The open frame was min structure for testing drive performance

No you are good with your clarity. But only having smaller scale design to base my opinion on, I simply pointed out my concerns. I think this has great promise and am excited to see what the FRC scale version is capable of.

[DavisDad]

http://simhardware.org/img/GearBox.JPG
http://simhardware.org/img/GearBoxSection.JPG

[magnets]

http://simhardware.org/img/GearBox.JPG
http://simhardware.org/img/GearBoxSection.JPG

This is cool, but I'm worried about the thrust loads the shaft on the CIM will receive. It's been a while since I've taken a CIM motor apart, but I'm pretty sure you don't want to be loading the shaft like that. If you placed some spacers on the CIM shaft so that any downward load on the shaft would be transferred to the plate the CIM is mounted on, you'd probably be better off.

[DavisDad]

This is cool, but I'm worried about the thrust loads the shaft on the CIM will receive. It's been a while since I've taken a CIM motor apart, but I'm pretty sure you don't want to be loading the shaft like that. If you placed some spacers on the CIM shaft so that any downward load on the shaft would be transferred to the plate the CIM is mounted on, you'd probably be better off.

Hey Magnets,

Yea, there are still some loose ends on the design. I need to check the thrust load specs for "deep groove" ball bearings. If they can handle the loads, the gear will butt against the inner race; otherwise we'll have to use thrust bearings. I didn't fix the gap between pinion and bearing yet.

[DavisDad]

Had to move planetary gear set outside wheel...

http://simhardware.org/img/GearBox_26dec13.JPG

[geomapguy]

This might be useful for you guys...

http://www.andymark.com/product-p/am-2622.htm

[DavisDad]

This might be useful for you guys...

http://www.andymark.com/product-p/am-2622.htm

Hey geomapguy,

Thanks for the AM suggestion. I looked at that, but want to optimize for FRC space and function. And, I'm learing about gearing; this is my first time building a gearbox. We may end up using a COT as the designs so far are more complex than I'd hoped. I haven't received the Matex planetary gear set (pic & link below) and need to determine the shafting requirements for alignment and loading.

Also, to use the CIM with the AM planetary set, you have to trim the motor shaft 1/4". Is modifying the shaft FRC legal?
.
http://simhardware.org/img/mainimage2.jpg
PN: 75-5MLD (http://products.matexgears.com/viewitems/all-categories/planetary-gears-torque-range-45-300in--lbs-?forward=1)

The 5:1 ratio units I ordered cost $73 ea

[DavisDad]

Progress update:

I've got 3 designs roughed out:


Planetary 5:1 reducer located in the Mecanum hub. I was excited about this when I found this video: e-Bike Hub Motor (http://www.youtube.com/watch?v=pH-52IgJui8) I immediately thought of the Sturmey-Archer (http://en.wikipedia.org/wiki/Sturmey-Archer) 3-speed bikes of my youth. I got an old unit from my local bike shop's junk bin and thought I had a cheap source for the planetary set. The design I've sketched requires pretty complicated shafting and I'm doubtful about putting the planetary at the end of the gear train (shock stresses). the current design requires disassembly of the Mecanum to access the planetary set :eek: Also, I envision an oil dip system for lubrication and expect this to be a problem in the hub.
Move the planetary set outside hub but on wheel axis. This has the same issues with gear tooth stresses.
Move the planetary set to motor shaft and integrate into one gearbox with the 3:1 bevel gear set. AM has a COT setup for this, as geomapguy pointed out: AM COT (http://www.andymark.com/product-p/am-2622.htm)
http://simhardware.org/img/am-2622-3.jpg


I haven't totally given up on the gear-hub idea. I'm waiting for the Matex set to arrive. When I get it in my hands and get a better feel of mounting/shafting requirements, I'll know better how to proceed...

[geomapguy]

Is modifying the shaft FRC legal?


Yeah it's legal.

Modifying the shaft and removing the window motor locking pins are the only legal modifications you can make to a motor.

[cgmv123]

Is modifying the shaft FRC legal?

Yeah it's legal.

Modifying the shaft and removing the window motor locking pins are the only legal modifications you can make to a motor.

Obligatory disclaimer that the relevant rules can change for the upcoming season.

[DavisDad]

Thanks geomapguy and cgmv123 for your replies.

I think we'll go this route; planetary in-line w/ motor and 90 deg bevel set to wheel. This is much simpler than the previous designs.

http://simhardware.org/img/GearboxPlanetaryHousing.JPG
http://simhardware.org/img/GearboxPlanetaryHousing-2.JPG

[DavisDad]

My son and I made these movies (links below) to show Mecanum with speed control. This is the FTC teams prototype; not FRC. This is the design from which we're scaling up.

http://i1.ytimg.com/vi/QL6rHh3uY8Q/mqdefault.jpg (http://www.youtube.com/watch?v=QL6rHh3uY8Q)

http://i1.ytimg.com/vi/UOgjS3hmQHI/mqdefault.jpg (http://www.youtube.com/watch?v=UOgjS3hmQHI)

[geomapguy]

Looking good.

[yash101]

The size factor is similar to my vex omnibot. With that size, you could use very grippy materials to get a good grip. The vex wheels are probably the most grippy wheels I've seen. The robot's ability to push is just like out 2013 robot. Have you tried to climb walls with it? (I mean going at the wall, full speed and scale it until the robot flips over)

[DavisDad]

...Have you tried to climb walls with it? (I mean going at the wall, full speed and scale it until the robot flips over)

Not on purpose. :)

[DavisDad]

http://simhardware.org/img/ConstructionDriveAssembly.JPG

http://simhardware.org/img/ConstructionDriveAssembly-2.JPG

[geomapguy]

http://simhardware.org/img/ConstructionDriveAssembly.JPG

http://simhardware.org/img/ConstructionDriveAssembly-2.JPG

So what all reductions do you have going on?

What's the mecanum wheel size?

Custom, VexPro, or AM mecanum?

[DavisDad]

So what all reductions do you have going on?

What's the mecanum wheel size?

Custom, VexPro, or AM mecanum?

Gearing: 1:5 planetary, 3:1 bevel = 15:1 total
Wheel is AM 8" MecanumAM Mecanum (am-2115) (http://www.andymark.com/8in-Mecanum-HD-Right-p/am-2115.htm)

[geomapguy]

You'll be coming in at around 10 ft/sec which isn't too bad.

Pushing match current is 51 amps which isn't great but not bad.

Overall, a very good design and setup.

[Ether]

You'll be coming in at around 10 ft/sec which isn't too bad.

Pushing match current is 51 amps which isn't great but not bad..

How did you calculate these numbers.

[geomapguy]

How did you calculate these numbers.




JVN

[Ether]

JVN

What values did you use for drivetrain efficiency and speed loss constant? And wheel coeff of friction?

[geomapguy]

What values did you use for drivetrain efficiency and speed loss constant? And wheel coeff of friction?




Darn didn't even think that those could be different.

I used .7 for coefficient of friction (AM 8" forward cof) since most pushing matches are in the forward direction.

81% speed loss constant
90% drivetrain efficiency
Robot weight at 155 lbs

Results:
Unadjusted Speed: 12.36 ft/sec
Adjusted Speed: 10.01 ft/sec
Pushing Match Current (per motor): 51.08 Amps

[Ether]

Darn didn't even think that those could be different.

There is no reason to expect those numbers to be the same for all drivetrains, as they are substantially affected by the type of drivetrain design (chain and sprocket, belt and pulley, direct drive, type of gearbox, etc) and workmanship (chain or belt tension, proper assembly and lubrication of gearbox, wheel alignment including toe-in and camber1, wheel axial offset2, etc).

1Toe-in and/or camber of a wheel causes the wheel sprocket (or pulley) to be non-coplanar with the driving sprocket (or pulley), and thus contributes to friction between the chain (or belt) and the sprocket (or pulley). Toe-in also causes scrubbing friction with the floor surface.

2"wheel axial offset" in this context means that the wheel sprocket (or pulley) is axially offset from the plane of the driving sprocket (or pulley), causing the chain (or belt) to be non-planar, thus creating additional friction between the chain (or belt) and the sprockets (or pulleys).

[geomapguy]

There is no reason to expect those numbers to be the same for all drivetrains, as they are substantially affected by the type of drivetrain design (chain and sprocket, belt and pulley, direct drive, type of gearbox, etc) and workmanship (chain or belt tension, proper assembly and lubrication of gearbox, wheel alignment including toe-in and camber1, wheel axial offset2, etc).

1Toe-in and/or camber of a wheel causes the wheel sprocket (or pulley) to be non-coplanar with the driving sprocket (or pulley), and thus contributes to friction between the chain (or belt) and the sprocket (or pulley). Toe-in also causes scrubbing friction with the floor surface.

2"wheel axial offset" in this context means that the wheel sprocket (or pulley) is axially offset from the plane of the driving sprocket (or pulley), causing the chain (or belt) to be non-planar, thus creating additional friction between the chain (or belt) and the sprockets (or pulleys).



I'd assume the way to calculate both are far too complicated.

[Ether]

I'd assume the way to calculate both are far too complicated.

You can plug in swag values and get ballpark estimates for a drivetrain you haven't built yet.

Or you build your drivetrain and then measure them. Then you can use the model to generate reasonably accurate data for that drivetrain.

[DavisDad]

Ether & geomapguy,

Great stuff! It didn't occur to me to model the system as part of the design process; I just used experience from previous designs. E.g., we used AM 8" Mecanums last year with the Toughbox Mini (am-0654). (http://www.andymark.com/product-p/am-0654.htm) with 10.71:1 gear ratio. I picked 15:1 as the bot was plenty fast at full speed in tank-mode but we had issues with mecanum control and I wanted to begin conservatively. Fast is fun but I think good controlability should be achieved first.

For those of you like me, who didn't have a clue what Ether and geomapguy were talking about ;) , here's a link to information that Ether has posted:

Drivetrain Acceleration Model (http://www.chiefdelphi.com/media/papers/2868)

This is excellent infromation!

http://simhardware.org/img/DriveTrainAccModel_Ether data.jpg

Getting a model calibrated should be very useful for developing an energy management strategy.

Now if I just postpone sleeping for 3 months, I think I could work through all the study materials y'all have given me...

[DavisDad]

Since we’re designing around the Andy Mark (AM) 8” Mecanum wheel, I’m going look closely at the assembly of the delivered kit.

I know there are issues with binding in the rollers if extra work isn’t done to prep the parts. I’ve looked at the video provided by AM: AM Video (http://www.youtube.com/watch?feature=player_embedded&v=Z9WbJ9atLxI)

I found the all middle rollers very sticky. This was not metal-to-metal interference but plastic or rubber to the brass shaft. I tried the AM video method but had no luck and don’t understand what it’s supposed to do.

I tried an 0.251” reamer to get a better fit (roller ID is ~ 0.0252”). This did not act as expected: once through with gentle rotation would reduce interference. What I found was rubber or adhesive like particles (no metal) on the reamer flutes (see photo below).

http://simhardware.org/img/MecanumRollerBidding.JPG

Here's an AM drawing of middle roller with my mark-ups of dimensions I measured (approx. avg.):

http://simhardware.org/img/AM-RollerDwgJPG

The metal bushing's ID is about 1 to 2 thousandths over the brass axle 0.250". The axles I measured have pretty good tolerances; eccentricity within about 0.0005". I don't think the binding I've found is a result of metal interference tolerances but from Polycarbonate
Core molding operation. I used a 3/16" reamer, inserted up to the far bushing and scraped between bushings. This method combined with reaming bushings with 0.251 reamer, resulted in a fit where the axle would fall through with no force.


I want to reduce the play in the middle roller and make the outer rollers spin better. See video:

http://i1.ytimg.com/vi/QKkLHd1Bwlk/mqdefault.jpg (http://www.youtube.com/watch?v=QKkLHd1Bwlk)



More to follow...

[DavisDad]

http://simhardware.org/img/ConceptualLayout.JPG

[geomapguy]

http://simhardware.org/img/ConceptualLayout.JPG

Is this designed for there to be a hopper or other structure in the open space connecting the two sides? If not, I'd recommend adding crossbars there as well for added support.

But looks good otherwise

[DavisDad]

Trying to use the space on the side and rear perimeter for electronics. The front is left open for game systems. The forward structure will be determined later by geometry of game systems. The wheel wells/gearboxes form the structural base for the box trusses on sides and rear ( truss bracing using 1/4" Al rod not drawn yet).

[DavisDad]

Changed gearbox to protect encoder

Previous:

http://simhardware.org/img/ConstructionDriveAssembly-2.JPG

Current:

http://simhardware.org/img/GearboxPlanetaryHousing-E.JPG

[DavisDad]

Received the Matex planetary gear set:

http://simhardware.org/img/MatexPlanetaryGearSet.JPG

Am ready to machine housing:

http://simhardware.org/img/GearboxPlanetaryHousingDWG.JPG

[DavisDad]

http://simhardware.org/img/HockeyBOT.JPG

[DavisDad]

Oh boy! This year's game is very interesting regarding drive design. We're still scratching our heads over game play strategy. At first read of the rules, I thought this was going to be a heavy contact Hockey game where chassis strength would be essential. I had images of robots checking each other, pushing and shoving to get/deny the ball.

Our Regulatory Affairs Officer ;) says there will be almost no defense, but I'm still optimistic...

From our initial brain-storming, it looks like a ground level opening in the front could be a real asset. I also think a low clearance/CoG has very desirable. If they didn't cost so much, I'd switch from the 8" Mecanums (already purchased) with 6" size.

Our drive team are leaning toward using the AM KOP drive chassis with regular wheels. I'm trying to steer them away from 4-wheel tank drive because of ugly turning and pivoting performance. But, I'll continue working on the modular setup as a design exercise.