Team 1658 prototype flying inverse differential(gyro-encabulator) bi-swerve

[LCJ]

This is pretty wild. Really cool idea and thanks for sharing!
What was your inspiration?

[sanelss]

Originally Posted by Chris Endres

I had a similar idea over the summer, but I never had time to design it. Looks awesome!

Are there any apparent problems? Does the differential work as planned?

we already did a test with half a robots worth of parts and it looked very promising so we decided to continue and make the whole bot. we already know it WILL work, it's just a matter of how well as an overall design

[sanelss]

Originally Posted by Ginger Power

I'm sure it's not optimized for weight yet, but what does it weigh?

in that configuration around 48 lbs however we can easily shave off over 5 lbs since a lot of the stuff on here was just thrown on because we had it and would work rather than trying to make it as low weight as possible. we will be well under 45 lbs on an actual game implementation.

[sanelss]

Originally Posted by asid61

This is the coolest swerve drive (or drive, really) I have ever seen!
As far as this versus a shifting swerve goes, an IVT only gets a maximum torque equal to the maximum torque at the lowest gear ratio, so lowering the speed does not increase your torque unless the motors are drawing more current (not sure on that last part). I really want to see efficiency/ acceleration specs on this beast.
You also won't be able to do turn + drive maneuvers without scrubbing, it looks like, but you hardly need those anyway.

actually you can turn+drive without much issue. it just wont necessarily rotate around it's center depending on the driver/rotation relationship

[sanelss]

Originally Posted by nuclearnerd

What algorithm are you using to choose the motor speeds? Lets say you want to speed up the right side. You could speed up the top motor, slow down (or reverse) the bottom motor, or some combination of the two. How do you avoid running either motor at an inefficient part of its range (such as stalling one while the other runs flat out)? Just curious.

still working on it but with the previous test we did my approach was as such:

the top cims are the reference motors, they always run just about full out but still PID speed controlled(so lets say 1k rpms) but that's completely up to us and arbitrary depending how much total power we want available. i will try to remove this though and just have them be full forward or full reverse without needing the extra control here. As you said there are a variety of ways to control this and we are still researching. Not sure how well that will work but will see. then the opposite side is variable speed to give us the speed we want from the output. with some simple math(just the average of the two inputs) we can get any output speed we want. so currently each cim is being speed(or position) PID controlled. We made half a bot work and currently in the process of making the whole bot work. Whole point of doing it is to see just how well and effective this design can be, it may or may not be favorable and that's why we're doing it, to find out!

[sanelss]

Originally Posted by LCJ

This is pretty wild. Really cool idea and thanks for sharing!
What was your inspiration?

a few years ago we had the idea of using an IVT on an FRC bot. There are enough advantages where it seemed like it would give us a very competitive edge in the competitions. looking online we found nothing though so we started playing with ideas. It took a while but we eventually came to realize that a differential could be used in such a fashion, but the obvious drawback is that it requires two motors per transmission and you would have to run them in active feedback speed control. You simply can't use that many motors and at that time we didn't have the capabilities to pull something like this of so it was put in the back of our minds.

So far we have tried various drive systems and the one we had the most success with is 4 corner holonomic which we used for years. It has a lot of it's own drawbacks though so we always had ideas on new drive systems and tried out different things over the years. We had to keep the maneuverability we were used to but we had to gain in speed and power. The only solution was some type of swerve. We prototyped a shifting swerve last year(of which a lot of parts can be seen on this bot) but when we found out the game had no defense we felt it was unnecessary to give up on the holonomic just yet. Due to this fact we were able to perform like we never had before since the weaknesses of holonomic drive didn't matter in that game and we were able to win the st.louis regional since we spent all of our time scoring rather than fighting other robots.

Well we know bumpers will be required again this year, which means robot on robot contact. From previous years we know that if you have a good scoring bot(which we had in previous years but would always get shut down) the other team will dedicate 1 if not 2 of their alliance bots just to get in your way and do anything they can to stop you. Well we don't want to be stopped. so we're trying to come up with the most effective solution we can think of. One of the issues with any drive system or motor control is you lack any control in the low end due to how the power is varied to the motors. so if you want to go slow or make fine adjustments most of the time you can't unless you over gear it which isn't desirable or spend a lot of time trying to tweak in overshoot/undershoot the best you can with PID control which in the end may still not be enough. So with this design we kill multiple birds with 1 proverbial stone.

As with any design however, it has it's own advantages and disadvantages. While we are aware of some of them already, we decided to build a whole bot to try to determine most if not all of the rest of them with such an implementation.

[GeeTwo]

Certainly different! I haven't seriously tried to wrap my brain around a swerve drive with pair-coupled wheel speeds and fully independent steering, but I have done a bit of thinking about the transmission.

An inverse differential is a great idea to work around the motor controller dead band, but I don't think this is the right application for it. In the no-friction approximation, a the speed of an inverse differential on the output is equal to the average of the input speeds, and (if frictionless) the torque is equal to the sum of the input torques. If you're driving the wheels in opposite directions, this implies that the speed is equal to half of the difference between the magnitudes of the input speeds, and the output torque is equal to the difference of the magnitudes of the input torques.

If I've done my mental algebra correctly, this implies that if the two motors are driven in opposite directions, the net output power of the inverse differential is never greater than you could have obtained with a single motor at that speed (leaving motor failure out of consideration for the moment). If so, putting two CIMs into such a drive system is just a better way to generate heat than putting two CIMs into an ordinary gearbox; they have enough thermal mass to handle that bit of back driving when you need low speed, low torque (single motor dead band) maneuvers.

It seems to me that a better application for the inverse differential would be for a manipulator arm driven by low-thermal mass, fan cooled motors. You could run them at (for example) 51% and -49% of free speed to get a 1% output speed, while providing plenty of cooling air to both. If I'm figuring things correctly, when under an appropriate load the reverse motor would be running nearly free at around -6V, and the forward motor would be running around 12V (at the top of the power curve), so you could get nearly half of one motor's stall torque at very low (or even zero) speed for a prolonged duration, making it great for holding up elevators and arms without brakes and without releasing the magic smoke.

[GeeTwo]

Originally Posted by GeeTwo

I haven't seriously tried to wrap my brain around a swerve drive with pair-coupled wheel speeds and fully independent steering, but I have done a bit of thinking about the transmission.

Update on this point: I haven't done any formal knematics with this yet, but if you're going to couple wheel speeds while having independent swerve steering, there does not appear to be significant benefit to having two different drive trains each powering two wheels compared to having one system that serves all wheels. A single drive train system with wheels in a rectangular or equilateral triangular configuration can translate in any direction or spin around its geometric center with no slippage. It can be steered in directed motion similarly to an automobile (or even a fire truck). It can also rotate around any point which is not too near a wheel with minor wheel slippage. It is also amenable to use with a 3-wheel drive system, which would reduce cost, weight, and number of parts compared to a 4 wheel system. The 3 wheel system would also have fewer places to avoid trying to rotate about, and wider avenues to transition the center of rotation from infinity to the center.

[AdamHeard]

Originally Posted by asid61

This is the coolest swerve drive (or drive, really) I have ever seen!
As far as this versus a shifting swerve goes, an IVT only gets a maximum torque equal to the maximum torque at the lowest gear ratio, so lowering the speed does not increase your torque unless the motors are drawing more current (not sure on that last part). I really want to see efficiency/ acceleration specs on this beast.
You also won't be able to do turn + drive maneuvers without scrubbing, it looks like, but you hardly need those anyway.

http://www.chiefdelphi.com/forums/showthread.php?t=2548

217 did something similar back in the day before Vex, AndyMark, etc...

Very impressive.

[asid61]

Originally Posted by AdamHeard

http://www.chiefdelphi.com/forums/showthread.php?t=2548

217 did something similar back in the day before Vex, AndyMark, etc...

Very impressive.

I remember 217's IVT (and designed a couple of my own) but I did not know they put it on a swerve drive! How well did it perform?

[IKE]

Cool prototype. The long unsupported chain spans may become problematic. Chains in that config like to derail. IVTs are very cool. Be mindful of 0 out put speed as some designs end up throwing a lot of power into the gearbox and cause damage. I am not exactly sure how your set up is configured, but would love to see a diagram.

[sanelss]

Originally Posted by IKE

Cool prototype. The long unsupported chain spans may become problematic. Chains in that config like to derail. IVTs are very cool. Be mindful of 0 out put speed as some designs end up throwing a lot of power into the gearbox and cause damage. I am not exactly sure how your set up is configured, but would love to see a diagram.

we already added one set of chain stabilizers/guides on the long runs which seem to have addressed floppy chain there but may need additional support on the diagonal runs. We will address issues as they arise. There will be several modes of operation with this design. A "direct" mode where the motors are always driven in the same direction and we just vary the pwm/speed(depending on control mode) going to the coupled motors. This mode won't give us any advantage of the IVT but good to have as a fallback. An "efficient" mode where we run the motors at their peak efficiency instead of all out in opposite directions in an attempt to conserve power, and a "power" mode where we go full tilt and will indeed dump all available power into it. in 0 output we just completely turn the motors off though. It's certainly still in the developmental and non field tested but we're doing what we can to get there. Hopefully before the new season starts so we can see if it's even worth implementing on a real competition bot.