4-speed gear automatic gearbox

I jokingly mentioned to a teammate earlier that we should cad a 4-speed gearbox, but now that I think about it, has anyone else ever done a 4-speed gearbox and controlled it like an automatic transmission in a car? It seems fairly straight forward, have an encoder on the shaft to determine if the robot is either stalling (trying to push another robot) or has stopped accelerating (has hit full speed), and in response, either shift up or shift down.

Has anyone done anything like this and do you think it is worth doing?

FRC0033, Killer Bees, did one many moons ago. http://www.chiefdelphi.com/media/papers/1580

Bear in mind that the motors used are no longer FRC-legal.

And yes, this was automatic. I seem to remember that associated code filters wound up in a sponsor’s product because they were handy.

Not sure if anyone has ever done it before, but from a practical standpoint, I can’t imagine a situation that would justify the complexity and relative bulk of a 4-speed gearbox.

From a design standpoint it’s relatively straightforwards. It would just be either the addition of an extra shifter stage (if you wanted to try to use COTS shafts/parts), or a custom shifting shaft and multiple-position cylinder/actuator.
If you wanted all four speeds on the same shaft, you’d probably find it easiest to use a custom ballshifter. (Though I don’t have any experience with >2 speed drives.)

Awesome. They managed to make it pretty compact. It doesn’t look much bigger than your avg single speed gearbox.

Was it ever built/used? Is there a video of it?

Team 968 did a 4-speed gearbox a couple different years. I want to say it was 2004 and 2005. I don’t remember if it was autoshifting though.

On 696, we tried an autoshifting algorithm on our 2-speed this year and did not find any appreciable distance in field-crossing time when compared to leaving it in high gear, even with varying the shift point through a range of set points.

One of our mentors made an excel graph of acceleration without friction with selectable speeds and calculus and stuff. It basically showed that speeds from 10-30fps will accelerate at almost the same rate until they top out at max speed. The only real change in accleration rate was noticed when we changed the inputs from 4 cim to 6 cim.

It’s not extremely difficult to autoshift. Normally you would be in high gear, but the driver would have a low gear override pushbutton to go into low gear for normal driving. Then you could just put a current sensor in the motor circuit and shift into low gear whenever current gets too high.
4 speeds seems excessive to me, because all you really ever need is a normal drive gear and a low pushing/ positioning gear.
You can also just use PWM control to lessen the duty cycle of the motors if current spikes.

I really do think that with enough work at it, this could be a better solution than a single or two-speed drive. The real problem that any team faces is shifting speed vs. the actual advantage gained from a new speed.

Side note: I believe that 217 did a CVT in 2002. That may be of interest.

Here is the main problem with anything more than a single or 2 speed gearbox:

The field is 54 feet long, and (this year) 25 feet wide. The furthest distance you could possibly need to travel, straight line, is ~60 feet.

The question is not whether you get more top speed out of 4 gears, but whether or not you can do such a thing in 60 feet, including stopping. If you have a straight line top speed of 30fps (or 1116.437fps for that matter), if you cannot reach said speed by the time you cover the length of the court, you have an improperly designed gearbox.

We ran a 5.5fps and a 15.5fps drivetrain this year, with 6 CIMS in ballshifters. The actual, tested, realistic acceleration time we got from starting in first gear, reaching 5.5fps, then shifting up “on-the-fly”, only gained us hardly measurable time. We, like most if not all teams who use shifting gearboxes do not use them for decreased acceleration time, but for torque vs speed selection-low gear for plowing, high-gear at all other times. Integrating some smart code to decide which gear to use based on wheel speed is definitely a good thing, but not why the shifters were chosen for the design.

Is such a challenge to build a compact 4speed transmission a great off-season project? sure- but consider this before using it on next years game: the field is everything.
If you save 0.1 seconds crossing the field with your new system, but you only do it 2-3 times a match, is it worth all that effort just for the extra 0.3 seconds you gain? Also, if we have another game like 2010 or 2010, having more speeds is useless because the field is even smaller.

Also, will you ever get a straight-line path? more often than not, no. There will be some field element, or one of the other 5 robots in you way, that you will have to steer around. Steering slows you down, in all cases of tank drive, as you must slow one side to turn that direction. That will further hinder your capability.

Sorry for being the Debbie-Downer, but that is my $0.02

1 Like

We used auto shifting this year, and it did help us get up to speed faster. I can’t comment on the specific technical aspect as to why, but in both 2013 and 2014 our robot always starts in 1st then shifts up to 2nd on the fly then back down to 1st when we come to a stop. It puts less strain on the battery - we could leave it in second all the time while we had a fresh battery, but found it could quickly get run down.

recommend less led lights on the robot next year 624, it might save some battery :] .

When I was on 195 as a student in '05 I copied the 33 design, made a few changes and we used it on our robot that year with our (at the time) perennial tank track drive. It was cool, fun, and a great experience BUT wholly unnecessary.

Not that this is the thread for that, but the amount of LEDs they had on their robot…I don’t even think they were drawing much more than an amp.

Here’s 217’s whitepaper on their CVT (from 2002). http://www.chiefdelphi.com/media/papers/1361

Team 1503 built an offseason drivetrain that utilized two 4 speed transmissions designed by Pat Fairbank. Check it out here.


It would be cool and fun to design, but probably not practical for an FRC game if used like an automatic transmission. Like other people have said, the field is too small to decrease travel time significantly.

Automotive automatic transmissions are used because of a narrow range of speed from the power source (engine) and a high range of output speeds. Electric drives typically don’t use a shifting transmission because a motor produces torque from 0 RPM to a high RPM plus they can remove cost associated with a transmission and final drive.

I could see a 4 speed more useful in FRC if used like an ag tractor trasnsmission where different speeds are used for different tasks. For example:

speed 1 - power takeoff
speed 2 - ultra-low pushing gear/high precision driving (4 fps)
speed 3 - low gear for typical game play and acceleration (8 fps)
speed 4 - high gear for quickly travelling 60 feet (16 fps)

I don’t think FRC needs a 4 speed transmission.

Based on the modeling work I did with shifting transmissions, I found that it was only beneficial to upshift automatically for improved acceleration if the low gear was very high, around 8fps for a 4 CIM 150lb robot, and this drove a desired ratio spread of <2:1 to put the high gear around 15fps. All of this resulted (in simulation) in optimum accel times, time to distance (short and long), and battery usage for pick and place maneuvers (which can increasingly use ‘low’ as the gear for short distance maneuvers).

However, for a 4 CIM drive, 8fps is not traction limited at 40a/motor with any reasonably grippy wheel. So that drove the desire for a 3 speed transmission, where 2-3 are used for ‘normal’ play and 1 is used for pushing.

Long ago, when motor power was significantly limited, shifting up sequentially would improve performance a lot (compared to throwing more power at it, which is common now). When the 33 4-speed was designed, this was the case.

Using dog and mesh shifting methods, it’s easier to combine 2 2-speed shifters with a different spread in series than it is to build a native 3 speed (with ball shifting a 3 speed is trivial to design, as you just add 1 more of the shifted gear sets). However, this results in a shift combination of L-L, L-H, H-L, H-H, where the 2<->3 shift is a ‘double swap’ shift. Automotive transmission engineers know that a double swap shift (more than one apply and release element) is a bad thing, and is extremely hard to synchronize properly. This can lead to temporarily being in a different gear (e.g. if shifting from 2-3 you can end up in 1 momentarily) which is really bad. So the answer then is to jump-shift, going 1-2-4 or 1-3-4. Then you get into some rather cool software, and a lot of design work to optimize all of the gears to get the performance benefit you are looking for. Definitely a cool project if you want an engineering challenge, but probably not worth it for an FRC robot.

So, basically, you go back and question why you wanted 3 speeds instead of 2 (or 1).

Now for some insight on automatic transmissions in cars. Modern automatic transmissions usually fall into 2 categories: Planetary wet clutch and automated manual/dry clutch.

Automated manuals are basically a series of dogs (or synchronizers, which are more complex dogs) each with 2 ‘gears’ and capable of shifting between 1, 2, or neutral. With many of these in parallel, any individual gear can be selected (1 at a time). The input is also coupled to the engine via a clutch, which can be released during a shift. Dual clutch transmissions have two of these pathways with separate clutches, and can ‘preselect’ a different gear on the other pathway then engage it by controlling the release of one clutch and the apply of the other.

Planetary wet clutch transmissions are basically a combination of planetary gearsets and clutches that either lockup (lock a segment to ‘ground’), input (lock a segment to the engine/motor), or connect 2 segments together. Virtually always, this is done with wet multiplate clutches, which can be slipped to smooth the shift. If you want to look up some cool gear layouts, look for Simpson or Ravigneaux type compound planetaries, those are basic 3+R and 4+R designs.

Team 190 also had a CVT in 2002, though we only drove about 20 foot at the start of a match before we parked it for the remainder.

Their “CVT” is actually an IVT. Most CVT and IVT designs are useless for FRC because their sole purpose is output speed, not torque. 217 is an example of a constant torque drive. FRC uses electric motors, so we don’t need an IVT/ constant torque drive.

To answer the first question, yes, I do believe it was built and used in 2004. (And apparently rather effectively.

As for the video… Zondag, IKE, Palardy, you guys know of any? I’m pretty sure most of the usual sources wouldn’t have any–video didn’t become a really big thing for a few years after that.

:ahh: Mind = blown.

Seriously geeking out here. Thanks for the explanation and insight!