2791 Versus 1714: A 2010 Robot Comparison

Leave a comment

I had a bit of an odd team situation for this year. For the first 8 days of the year, I worked with 1714 as they went through their game strategy and initial concept design. After that, I went back to school and met with 2791, having already picked their strategy. It was a unique opportunity to see two different approaches to the same problem. Neither team collaborated and I didn’t tell 2791 what 1714 was doing or anything like that.

Keep in mind I have a bit of a case of “More Robotics fog”, so I may misremember key details of 1714’s design process. 😛 I also don’t know that much about the competition robot other than what I’ve seen and asked about. I can tell you they had shifters and a vacuum but I can’t tell you what was used to machine exactly what.

2791 on the left, 1714 on the right.

Strategy

Both teams leaned toward a midfield oriented robot. Oddly enough we had very similar strategic analyses as I left 1714. We figured that “any random robot” could be put in the front, and that a lot of teams would be able to inaccurately clear out of the back zone and then “defend” for the rest of the match, so the focus would be on a midfield bot that could control balls and leave the zone if needed. The midfield bot would be the most versitale for qualifications; it could start in the back zone and advance if we have to. It could advance to the front to clean up if need be. In the eliminations, grabbing a back zone 3 and D and a front zone cleaner would be very easy to do. The obvious choice would be to play the midfield first, as that is where the balls go.

We determined the bare minimum of teams that changed zones on any alliance would be 1. Since this was in our minds a requirement to beat any given alliance, it became mandatory for our teams to be that team that changes zones.

The midfield strategy we thought would call for a lot of pushing. Since the balls in the game would all return to the same spot, we all would have to fight for position under that spot. As a result both teams quickly decided on high traction drives. With the game encouraging defense in a lot of ways despite the seeding system, a mecanum drive just wasn’t an option either team wanted to pursue. A serious discussion of the swerve drive was had on 1714, but because the prototype wasn’t quite ready for prime time even on a flat field, it was abandoned.

Differences in Design

Right here is where 1714 and 2791 diverged. Oddly enough, it’s when I started working with 2791. 1714 built a 6 wheel prototype and figured out a configuration that worked over the bump. 2791 went for a simple 8 wheel configuration apparently using some CAD to figure out how 8 wheels would react better on the bump than 6. As I left, 1714 was working on roller and vacuum prototypes, but 2791 heard from 1726 that the roller magnet was relatively easy so we didn’t think too much of it. Whoops. 1714 made some kind of cool kicker that used pneumatics to pull back or something, I still don’t really get it. 2791 used a pneumatic precharge that kind of worked. 1714 shifted, 2791 wanted to but couldn’t afford it.

Hanging

1714 and 2791 both strongly considered hanging, but neither did in competition. After I left, 1714 decided that they could score 2 or more balls on the ground in the last 20 seconds, so hanging would be a waste of time. 2791 thought the consistent points that hanging offered would be invaluable in some matches, so a large focus was put on it. However, no hanging mechanism 2791 made worked in competition, within weight.

I was “on the fence” between the 1714 and 2791 decisions all year. Makes sense, I guess, being “on the fence” between the teams. On one hand, hanging was worth it. If Shaker hung reliably in addition to what we could do, we would have been a force at CT without a doubt. Few teams not named 469 scored 2 balls in 20 seconds on average, and a well designed hanger could go from normal configuration to points in 10. On the other hand, hanging took a ton of resources. Shaker’s ball game suffered a lot because we spent so much time working for those two points. Initially I thought it was silly that 1714 didn’t pursue hanging because I thought they didn’t think it was worth it, but now I see the wisdom of what they did. Doing “one thing well” is always better than two things badly.

At Competition

Both teams eventually shifted to a primarily front zone role. 1714 found that they were much more effective in the front though they could play the mid zone as well. However their consistency in front was their biggest asset. 2791 found that their kicker kind of sucked at getting balls over the bump consistently, and with last minute vacuums and lexan guards they played the front zone quickly and decently, so they switched to only the front zone. Both teams made the same strategy changes despite different looking robots.

1714 had a great run all year, starting out at Wisconsin as the first pick of Wildstang. Ultimately they were bested in the eliminations; Wildstang had to juggle playing both defense and midfielder, and 1714 was subject to HEAVY defense that made them take longer to score the balls in their home zone. Since they had balls in the home zone, and since 111 was defending, no one really contested the middle and ultimately they were narrowly bested by a strong 1732 / 2574 / 171 alliance. Then they went to Minnesota and dominated the event. There’s no video so I don’t know anything about their run, but they were pretty clearly the bets team there from what I’ve heard. At the Championships, 1714 reached a milestone and was selected for eliminations for the first time ever (1714 just missed the cut in 2009, and in 2008 they got carried to the 7th seed), pairing up with 1305 and 230, only to be topped in the semifinals by the potent combination of 1086 and 217.

2791’s run was less great. At WPI, the robot had a ton of problems and several “nightmare matches” threw us in the odd situation where we did so badly we seeded in the top 15 or so. We ended up throwing a few matches versus 230 (the event winner) and worked to rebound on Saturday. Two great matches and we got the 7th seed and 5th captain, picked what was left in the barrel after all the good midfielders left the field by pick #2, and then fought hard against a similar alliance, losing only after 5 matches.

CT turned out a lot better, as Shaker was focused from the beginning on doing what it did best: playing front. We swept our way up the qual match ladder, winning several matches, then faced the “Friday Afternoon Curse” that always troubles us, including a 1v3 match against 1501 and 1124 that we didn’t throw. After some strong matches Saturday including setting the regional high score with 383 (until the QFs…), alliance selection came and we were the 5th captain again. 177 and 173 were still on the board thanks to the latter’s failed Classmate all Friday and the fact that somehow people still don’t get that picking 177 is a great idea no matter what. We broke into the Semifinals, but not before technical difficulty after technical difficulty; 177 took a hit that immobilized their drivetrain, 173’s arm snapped repeatedly, one of the Blue Alliance’s DOGMA counters was finicky, 177’s hang missed due to a twisted cable, Shaker’s vacuums were finicky… you name it, it happened. Anyway we ended our season here since we didn’t have a Championship qualifier; at least we ended it with a fantastic alliance.

Who was better off?

1714 without a doubt.

Advertisements

FRC 2009: 1714’s “Philbert 2”

2 Comments

Edit: Check the post Collin made in the comments for fact corrections.

For my senior year I was fortunate enough to join a really great and well run FRC program, More Robotics. Here’s “where it all began” and where I got inspired to pursue engineering as a career instead of typing all day as a CS major. I did not do that much mechanical design on 1714, as I had better teammates, but I did help with the game strategy and ultimately got hooked around competition time. I was very much in over my head, and I called the game exactly wrong (continuing a streak that only ended in 2010), but it was a great year for me.

Anyway, I might as well show you the bot and stuff.

More Robotics 09

It kind of sort of looks like HOT before they made mods and became the best bot ever.

It’s a bit hard to visualize the robot because of the entirely polycarbonate construction, so here is a picture with some balls.

Here are some balls.

Here are some balls.

Basically, if HOT’s robot were one ball wide, it would work a lot like 1714. The front hopper holds about 3-4 balls gravity fed and the belt system can hold 6 balls if carefully managed (which it was, our operator was pretty good). The main design was that we wanted to be able to shoot from long range, on moving targets, quickly. We also wanted to avoid jams at all costs, hence the single track of balls. The all polycarbonate construction is because we worked out of our awesome sponsor’s plastic fabrication shop, American Acrylics. The robots also look fantastic and cool this way.

I’d like to emphasize I didn’t do much design work here, so any “wow you guys rocked gg chris” comments should be directed toward the other students of 1714.

Building with polycarbonate

Before 2010 (when 1714 got a CNC machine for LARGE sheets of polycarbonate), we got really creative with our fabrication. We primarily build with 1/4, and 3/8ths polycarbonate sheet, using 1/8th for shielding and non structural parts of the robot. Polycarbonate can easily be cut on a table or band saw, can be drilled into like most parts, and can be routed out like wood. Like sheet metal, using it in the right application can make for a decently rigid robot, and with Lunacy a rock solid metal frame wasn’t a huge priority anyway. So we made everything out of it.

The chassis is made up of 3/8ths polycarbonate sidewalls, all bolted together, with 1/8″ top and bottom covers for electronics. Pretty simple stuff. Some lightening holes here and there. Occasionally we’d make a bearing block out of some 3/8ths material and shove a bearing in there to support the shafts. We used the BB planetaries from 2007 with a supported shaft for the back wheels, and window motors with a large overgear in the front. The idea was that if the front wheels were spinning fast enough to “break loose” we could turn faster… I don’t know if that actually worked very well, but no one uses the Window motors anyway. We installed encoders to match the speeds of the wheels, since it’s not like either were torque limited.

The cool part was the ball tower. We needed a large load bearing structure but with a large lightening pattern. We had a laser cutter and lots of 1/4″ thick acrylic, so we prototyped with that and using some wood as spacers (where the polycarbonate “standoffs” are now). Once we got a design that worked, we cut the exact same hole pattern out but with huge filleted square sections all over. We then taped and secured the acrylic to 1/4″ polycarbonate, cut out a rough hole in the polycarbonate for each lightening square, and then used a hand router for the rest. It was a REALLY cool process and it made for clean, precision cuts. The other two sides of the ball tower were made with some 3/8″ lightened polycarbonate, and then we had a superstructure. (Can you tell this is the part I worked on a lot?)

The turret was the creation of our manipulator captain. The turret base was laser cut UHMW Polyethelene, which is fairly self lubricating. A large gear profile was cut in one of the layers of the turret; this gear mated with a smaller laser cut gear mounted to a Globe motor in order to rotate the turret. It was a simple, fairly low friction solution with the minor downside of making it hard to mount a vertical wheel in place. Instead two horizontal wheels were used for a bit more range, coupled to a CIM with some Poly Cord (polyurethane belting). We couldn’t “curve” shots but we had a turret and doubling the wheels used made our exit velocity pretty high. Before we modified the robot, we could make 20+ foot shots. Not accurately, mind you… We threw one of the shitty Mabuchi motors on a little “kicker” to get that last ball out of the turret and into the shooter.

Design Roadblocks

Our season had a lot of troubles. We initially pursued an Archimedes screw based design. Now a lot of teams say they have Archimedes screws, but actually have helixes or whatever. We pursued actually building an Archimedes screw, with the path spinning and everything. Unfortunately we failed to notice 1726’s cool videos online with the screw’s inside spinning, but not the outside rails the balls roll up on! Whoops. We scrapped the screw and went with the belts.

Our shooter was great… at long range. What wasn’t great was camera tracking. We were kind of counting on tracking goals and shooting long range. Nope. We also had traction control issues. Mainly instead of a PID loop, we had a fairly laggy system that would ramp up to the input specified, then check the input only after that. We also didn’t check if the wheels spun too fast at all, so we had not a very good system! We basically spent the first regional driving around and being a target. We missed eliminations entirely.

Oh, another problem: At short range we’d be running the motor incredibly slowly, but even at some of the lowest PWMs we had the shooter was still shooting a bit far.

Anyhow, we fixed both those issues with the addition of an AM Planetary to the CIM as well as new code with zero traction control whatsoever! We lost shooting range we no longer cared about and became a turreted flumper.

Competitive Success

So, yeah, did you know I won 1714’s first regional? 😛

I’m just kidding, I was just the drive coach so I didn’t actually accomplish anything. But yeah, 1714 went from missing eliminations to the first overall selection at Minnesota. While one could take that as a snide remark regarding the strength of the Minnesota Regional, we actually improved a lot. In one of our best qualifier matches, with the three best bots in the competition on one blowout, high score alliance, we scored 23 balls. Granted, it’s because no one in Minnesota bothered to starve balls, but still. We got better. I got a bit better at communicating to the drivers, the drivers handled the upgraded bot amazingly, and with the entire team behind us we got 1714’s first regional win. It also was one of the last regoinals won by a then-rookie alliance captain, 2826.

The reasons we managed to go from terrible to competitive are twofold. First, continuous improvement. We saw what was wrong with the robot and changed it. Second, we were fast. Our belt drive was coupled to a CIM and Toughbox. Our belts were moving at around 6 feet per second. We scored our starting 7 in a second. Because we could score so quickly, it helped to make up for our low capacity and ball output. Who cares if you’re a double wide dumper if you can only spit out 2 or 3 balls a second? Speed is EVERYTHING and this is proof.

It’s too bad we didn’t build a dumper, huh.

Things this design taught me:

  • Whatever you do, do it quickly.
  • The shotgun approach is sometimes the simple and elegant solution.
  • Asking for help isn’t a bad idea. What if we sent 1726 an e-mail?
  • Don’t rely on a solution you don’t know will work (camera tracking was SO reliable this year…)

Conclusion

This robot inspired me. 😀