Team 842's drivetrain for FIRST Stronghold!

[Donut]

Originally Posted by mrnoble

-this will be our first year ever traveling to Flagstaff. Is this the level we should expect to see, in terms of design and game savvy, from all the teams there? If so, I am a little bit excited, and a little bit scared. I've heard you guys are vicious as a region.

842 is one of the elite teams in AZ, there's a reason they won the only regional in the state 2011 - 2014. There are some notable other teams that will be in attendance at Flagstaff (125, 192, 2122 from out of state, 60, 2486, 4183 from AZ*) but 842 along with these others definitely represent the cream of the crop around here.

*I'm not familiar with any Colorado teams that are attending and have limited knowledge of some of the rising stars in AZ that have come on in the last few years when I lived out of state.

[mrnoble]

[Joe Johnson]

First of all, I love this chassis. Let me get that out of the way. Looks very nice. I suppose you're in for a ton of awards which you justly earned even if you guys do nothing more than just put bumpers on this chassis and compete with it as is.

Here are my concerns. I don't want to rain on anyone's parade. I am just trying to have a real discussion as though this was something that our team was designing themselves.

1st Comment
If you slow down the video, you can see that the chassis is flexing a bit as it turns. That long arm on the single trailing arm suspension and also the spacing between the dual rocker suspension give the wheels a big moment arm when they scrub over the carpet. That in turn puts some high loads into that chassis. I suppose that 842 has done some calculations to design the reinforcement plates. I just worry that over a season either the rivet will work loose or that someone will T-bone them against some field element where one of those gorgeous yellow tires will get "tripped" and then you've got the full push load of another robot trying to pretzel your robot.

2nd Comment
You can see the aluminum bits coming into contact with the defenses at a number of points in the video. It is not catastrophic because the robot clearly gets over the defense, but it DOES cause the robot to momentarily stop as a few points in the video. Why didn't you lower the wheel axles with respect to the aluminum bits so that this was less of a problem?

3rd Comment
When you have suspensions. Springs and torques from the drive belts and tire/carpet forces matter. Sometimes they matter A LOT. And they are not so easy to figure out (in theory they are but in practices, it is easy to screw it up and not consider all the loading cases).

Looking at the photo, it seems that the dual axle arm is completely unsprung while the trailing arm has a gas strut.

I don't have all the dimensions of things but my back of the envelope calcs (and scaling some photos which is not the most accurate way to get dimension, I assure you) show that the dual arm loads shift 10-20% depending on what direction the wheels are driving (said another way the normal load on the center tires is equal to the normal load on the front tires plus/minus 10-20%). This is probably not going to affect things during a pushing match -- I was afraid it might lift the entire robot but that seems unlikely unless the tire gets traction much greater than a CoF of 1.

BUT... that trailing arm is another story. Again, I don't have all the right dimensions and I don't even know the spring output or travel but the back of the envelope calcs estimate that the when pushing forward, the normal force of that rear tire is going to go up by as much as 50% (when pushing the friction on the carpet tends to lower the wheel, making it bite harder into the carpet). This could kinda be a big deal depending on the rest of the robot and the situation.

4th (and final) Comment
Suspensions are designed to keep weight on all the wheels. All the wheels. That means that the front and the back wheels have weight on them when turning. If you play the video you can see that this robot is not having a great time turning. I know I am like a dog on a ham bone but really all ya'll should read Dr. Joe's Seven Stages of Dealing with a 4WD Robot with Grippy Tires

So... ...842, hats off to you. Very cool. But there are a lot of details to making suspensions work and work well. Think hard before you decide to follow them down the suspension path.

Callin' 'em as I sees 'em.

Dr. Joe J.

[s_forbes]

Joe, good comments, these are all things we've been discussing and worrying about since we started trying to design the system. It's not an easy thing to build!

Just to clarify on our strategy for using this drivetrain, we do not plan to use the robot in a "suspension" mode for a majority of the match. We intend to lock the rocker mechanism into a "dropped center wheel" mode while driving on flat carpet and switch to the "suspension" mode only when crossing defenses (ideally we'll just be driving straight over defenses, so we won't be dealing with the scrubbing issues while turning like in the video).

With that in mind, here's where we currently stand w/regards to your comments:

1- We were anticipating the flex, but didn't have the resources to to accurately model it... It was faster for us to build this prototype with the components we were planning to use for the final design. The brackets shown are not final, they were made to let us do some testing and find the weak points of the design. The final bracket designs have some modifications based on what we learned, and lots of the high stress locations will be using bolted connections rather than rivets.

2- The frame contact (from the rocker) is something that was hard to work around using the components that we wanted to use. We saved a lot of manufacturing and design time by using Vexpro products, and tried to make the best geometry we could with them, but the dimensions of the wheels/versablocks/idler pulleys made it tricky to offset the wheel axles from the pivot location without making fancier brackets. We are still getting used to our new CNC, so keeping parts simple to speed up production was one of our priorities.

For our first tests of the concept, we literally just drove 39's robot from 2010 over our practice defenses to see how it would perform. It also had contact issues with the suspension members hanging up on the defenses, but it didn't seem to deter it (more importantly, it was still smoother than any other drivetrain concept we had tested at that point, despite the frame members contacting the defenses). It's not very pretty, but it still meets our requirement to cross the defenses in a controlled manner.

3- The torque provided by the belt on the rocker is a neat dilemma. Team 39's 2010 robot was designed so that the rocker mechanism is powered by one gearbox (attached to the rocker) and the trailing wheel is powered by one gearbox (attached to the trailing arm). This gets rid of that issue.... but we had decided pretty early on that two speeds would be pretty necessary for traversing the flat midsection and safe zone of the field quickly while maintaining the ability to push and drive over defenses in a controlled slower gearing.

We opted for the two-speed transmission with power transferred through an idler pulley at each of the suspension joints. Simple calculations showed that it would be able to push over the worst case defenses (with the rocker in front, not so much the other way around). The case we didn't look at was when the robot accelerates on flat ground in high gear, with the rocker at the back of the robot. The added torque on the rocker provided by the drive belt makes it pull some very large wheelies. This is another issue that is resolved by locking the rocker into 6wd dropped center mode.

I will see if we can run the suspension up against a wall in both directions with a scale to see how it reacts, that's a test we haven't done that should be interesting.

4- Shouldn't be an issue when we are locked into the 6wd dropped center mode. We never even tried to put it in high gear with the setup in the video, we knew it wouldn't turn.

[IKE]

Reminds me a bit of this Chassis, though a lot more articulation.

[jodge1706]

I would have missed it if it wasn't for all the comments about the end

[Joe Johnson]

Originally Posted by s_forbes

Joe, good comments, these are all things we've been discussing and worrying about since we started trying to design the system. It's not an easy thing to build!

Just to clarify <snip>

I love this about The FIRST Community and the ChiefDelphi Subset of that community. It is a great place to have a conversation about things we love. We are strong enough to have discussion where we share what would be closely held secrets in other contexts, folks think, questions form, questions get asked, more thinking, more discussion, answers come, more questions, more sharing, more answers, ... ...we all get smarter.

Hats off to 842. You guys are doing so many things right.

Cheers,
Dr. Joe J.

[s_forbes]

After having gone through 2 regionals, here's an update on what we learned with this drivetrain.

First thing to mention, I never showed our method of locking the drivetrain into 6wd mode. This was done by using a 4" stroke, 1.063" diameter cylinder that pulled a line connected to the front of the rocking mechanism. When activated, this forced the middle wheel down into the carpet with just enough force to allow it to lift the front/back wheels into the air, effectively putting all of the weight of the robot onto the center wheel. We used a Spectra line with an adjustable eye-splice on one end so that we could vary the length of the line, effectively making our "center drop" adjustable. You can see the cylinder and line in the picture below (click for large). Rubber bands are used to keep light tension in the line.



I think it's important to note that our drivetrain would have been a disaster without this feature! The suspension evenly distributes weight among all of the wheels, which is good for crossing uneven terrain, but makes it nearly impossible to turn on carpet. Toggling to a 6wd mode lets us drive like normal on flat ground.

During our first regional, we ran into some issues with the belt tension and wheel alignment when we went into the elimination matches. We hadn't been maintaining the drivetrain during qualifications, and it caught up to us. Each wheel is held on with a Versablock assembly, and by the time we got to eliminations many of the Versablocks had worked themselves loose. (by "loose" I mean "why is there only one bolt holding this on???")

A fun result of loose bearing blocks is misaligned shafts, which lets the belts put some side load on the pulleys, which pops the flanges off. I recommend avoiding that.

At our second regional (Vegas) we corrected the alignment and tension in all of the belts and kept a close eye on them, and never had any issues. I was actually surprised at the beating the drivetrain was able to withstand, as nothing worked loose or bent after our driver repeatedly rammed the robot over defenses at full speed. It can take much more abuse than I expected!

We do hear occasional belt skipping when the drivetrain is under heavy load (full reverse when going full speed forward, for example). I believe that the serpentine belt on the rocker mechanism is able to deflect the shafts enough to skip over the idler pulley. It runs over 5 shafts in total, and each is cantilevered, so there is quite a bit of deflection under high load. This is the portion that I think would need the most improvement in the next iteration.


Overall, I'd rate it 10/10 based on entertainment value, but 5/10 on effectiveness. Probably wouldn't do it again, but it's nice to have around.