Designing to Improve

After coming home from St. Louis, one of the biggest lessons I’ve learned is that it often helps a team’s competitive edge is they continue to add and improve upon their robot as the season progresses. While we were a consistently strong team at Finger Lakes, not needing any fancy gizmos to win the event, setting wheel onto the Galileo field proved to be a significant challenge for us. We were a moderate cycler, making 3-4 trips with about 80%-85% accuracy, but we lacked a floor pick-up or powerful climb that would help us stand out.

I am hoping to find some input from teams who have made significant improvements (I’m not talking tweaking some software or a minor addition; I’m thinking of teams who have added whole new mechanisms or redesigns in the past) on the following subjects:

Were the things you added part of your original design? Did your initial robot concept begin with all the bells and whistles, and you built what you could in the initial 6-weeks and added what you needed to later? Or did you build a simple robot and then added what you wanted after that based on ideas you picked up later?

Do you intentionally leave spaces on your robots where improvements could be added, or do you find space when you need it? If you don’t specifically have improvements in mind for the originally shipped robot, do you try and keep the weight/cost as low as possible to prepare for future changes?

Do you have a full or partial practice robot to test your changes? Do you have a practice robot at all?

After adding a mechanism, how have your results varied? Did they often come out working as you had intended, or do they require further tweaking before you begin using them consistently?

What other practices do you keep in mind to help with the addition of new mechanisms? Do you try to keep the entire design modular so subsystems can just “snap” on and off? Anything else?

I appreciate any and all’s inputs on any and all of these subjects.

-Leeland

We’ve only done minor additions but we did a few things this year to make those as easy as possible to do.

What really helped us this year with additions was adopting a bolt pattern. We really liked 973’s #7 drill .5" spacing bolt pattern. There are enough holes to easily find a spot for all additions and the #7 holes allows us to use 10-32 bolts or 3/16" rivets. This gives us a lot of versatility. We use 2"x1" tubing, so we only had the bolt patter on the 1" wide face this year. If we had to do it again we would add a bolt pattern to the 2" wide face too.

This year we bagged a cycler with no intake. Also, we initially had a passive climber. Lastly, the shooter got replaced for champs(to increase rate of fire). We decided to leave a place to attach an intake, and already had gearboxes for it designed and manufactured in case we ended up having time to add one. However, this addition never happened. Going into our first regional, we knew that we wanted to be able to go under the pyramid and and hang after the buzzer, however we didn’t have time to replace the passive hanger. It worked fine for us(though we had to be careful or the bumpers would get stuck on the ground), but our bolt pattern allowed us to easily add the active hanger for our second regional. The replacement shooter also mounted onto the bolt pattern. We like to leave weight for additions, but both weight and BOM were not an issue for us this year. BOM was significantly under and our robot only weight 80lbs as of bagging(eventually 90 lbs).

Unfortunately this year we did not have a full practice bot to test our changes on. The lifter never made it onto the practice bot. The weight of the two robots were significantly different, and we were sure it would work anyways. The replacement shooter did make it onto the practice bot, and this helped us a lot in tuning.

We didn’t have to tune much this year once we modified the comp bot. The lifter bolted on and worked perfectly. The shooter required a bit of tuning, but bang bang took no time at all to tune. We ended up losing consistency of our shooter later into champs, but I believe this is unrelated.

The last important thing to do to make mechanism go on and off really easily is electrical. This year, we soldered no connections. Instead, we found crimps that slid onto the motor terminals or used powerpoles. This allowed the superstructure and all mechanism to pull off the robot super easily.

After our less than desirable performance against you guys in semi’s at FLR we added another 18" to our FCS height and performed much better at Boston. We traditionally improve a ton between our competitions each year. I’ll highlight what we’ve done so far:

2011: New acquisition claw custom machined out of metal with plastic pulleys (CAD on our website)

2012: Pneumatic brakes to improve double balancing ability and co-op

2013: Shooter was raised 18", climber was removed due to lack of use.

Overall, we try to just design the robot the best that we can before our first competition and then use our 30 lbs very creatively to redesign the part that needs it before the next competition.

Our shooter and climb mechanism were both modular.

For our first regional (Palmetto), we bagged the robot with a version 1 climber and a working shooter. We brought a modified ascent mechanism to the competition and swapped it out (6 bolts). The ascent mechanism worked perfectly, but our shooter had a tendency to jam.

For our second regional competition (DC), we performed a high-risk shooter swap in the pit. We had fabricated a wiring harness on our practice bot, but not on the competition bot. This was a risk because we had to remove the working wiring harness and place a new harness in the chassis. Unfortunately, or modification of the shooter did not solve our root cause: servo motors acting as a gate for a gravity disc feed. We were not actively controlling the discs (bad idea).

At the Championship event, we again swapped our shooter for a full-on iteration of the feeding mechanism. It worked! We finally fixed our jamming issues and we performed nearly flawless (aside from some operator error and a rough impact with an opposing alliance).

The takeaway from this is to know what devices you may want to iterate and build a design that accommodates it. Incorporate a method for simple removal, use harness connectors on your wiring to the digital sidecar for each module, and use a standard, pre-determined/machined bolt pattern.

Oh, and use Anderson Power Poles. You can thank me now :wink:

After the SBPLI regional we scrapped our 3 level climbing system and at Chesapeake we installed new sides with a cut-out to enable a static, drive-on climb. It worked 60-70% of the time though the 10 points were never needed to win.
We are constantly trying to improve-and not just to win. We have a motto, “stop tweaking when you want to stop winning”. We tweak and change things whenever we can. In fact, just before the eighth Qual. match in St. Louis this year we did it again. We knew we wouldn’t get picked, but you learn to make changes under less than ideal circumstances and this makes your team better. We think it pays off in the long run because you learn to constantly re-evaluate and analyze your robot to make it better.
Last year there was a thread on CD which asked, “What separates the elite teams from the rest” --or something to that effect. The most poignant response came from an elite team. He essentially said, “Most teams try to shoot the ball the best way they can, we [elite] try to score the best way we can.” The subtle difference is in the end goal–scoring. Elite teams, I gathered, come up with the best method of _ _ _ _ _ _ _ -to win the game, then they re-evaluate it and improve upon it and keep doing it until the season ends.

At the beginning of the season we had a non-working 30 point climber which took up the entire front of our robot, and a fairly accurate shooter, but was very difficult to load because we had to load from the side. We also got stuck on disks because we had many things below our chassis. At champs, we cut off the climber, removed everything from below the chassis, added a pneumatic 10 point buzzer beater hanger. We also added a front loader, and our performance went up dramatically. We were able to do 4 cycles easily, vs the 1-2 we could do at our regional.

Did we design/plan in advance?
Yes and no. We had some bells and whistles planned that didn’t get done by Stop Build, some of which we worked in the withholding allowance, some of which we bagged. We also had several changes that were not planned beforehand.

(How) Did we leave space?
Yes, though “leaving space” is probably not a term I could apply to anything we did on a 28x28x28" robot. We work on modularity overall (see below), and try to estimate what we might want to change and how we could best do it. Keeping weight down is definitely a goal for this reason, though it’s not always successful.

Practice bot?
Yes, and we wouldn’t have even attempted to climb without it.

Does it work seamlessly, or is there tweaking?
Between both. Some go on without a hitch, others are a constant pain, tweaking throughout the event. (Or the next event. Or Einstein.)

What else do you do?
Be modular, have a practice bot, CAD everything. We try to think hard about what we’re really going to change and how to make it simple:

2013

  • Swerve modules switch in 5 minutes (4 bolts and 5 wires)
  • Swerve steering motors in maybe 10 minutes
  • Entire shooter comes off with 3 bolts (ok, 2 wires, 2 hoses and a quick release pin)
  • Shooter strut is entirely adjustable and can come off with another 3 bolts (of course, it does have the PDU on it…oops)
  • Climb wheels switch in under 10 minutes, though that’s basically the only quick thing about the climb
  • Arm, which we iterated a bunch of times and eventually tabled entirely, came on and off by punching out a single (very tight) axle

2011

  • swerve modules switch in 10 minutes
  • claw switched in something like 10 minutes; - arm came off with 4? bolts
  • minibot deployment was very modular and quick, 2-10 minutes for different things
  • minibot of course came off–got iterated a lot

Note I missed 2012, in which basically only our swerve modules were modular. You’ll note that by some odd coincidence, we won for our minibot and DT in 2011 (reasonable worlds performance for our claw). In 2013, we won for our drivetrain and shooter (some for the climb). In 2012 we won solely on our DT. None of these things were working at our first event as well as they were working at the last event we won. Weird happenstance, huh?

The robot we had playing on Einstein shared only its drivetrain with the robot we entered in our first regional.

At Central Valley Regional, our shooter was a 2-wheeled linear system using BaneBots wheels; our intake was a metal pan with rollers on top. We had a passive 10-point hanger that popped up and engaged when we rammed the pyramid.

By the Sacramento Regional, we had switched the shooter for a single-wheeled, J-hook design using a urethane drive roller. It gave us much more consistency than the previous system, and didn’t need to have its wheels replaced all the time. In addition, we changed our hanger to an active system with two beefcake pneumatic cylinders traveling up and down. This was due to the previous system’s tendency to get stuck when the back of our frame wedged into the ground.
Side note: at this regional, our shooter began the day not performing up to spec. When testing it, we had run it directly off a battery; when powered through a speed controller, it was much weaker for some reason. Because we live in the same town as the regional’s location, we spent Thursday night making a new gearbox for the shooter by hand, incorporating VEXpro gears we had lying around. It worked pretty well, surprisingly.

At Championships, we replaced our pan intake with one that dragged the frisbees along the ground before flipping them into our robot. It gave us much better pickup ability than the previous one while also being sturdier. Despite that, we bent its side plates a couple of times, and had to reinforce it with extra aluminum plate. We also changed our hanger again, this time attaching the hooks to flexible steel rods on top of the cylinders so as to not bend the pistons. Finally, we replaced the shooter gearbox with one that was not made by hand.

tl;dr Constantly iterating our robot made it much, much more competitive over the season.

Is this not prohibited?

Wouldn’t it just be in the withholding allowance? (if it met the total weight, of course)

EDIT: do’h, Thursday is one of them days you regional guys actually have an event on. Good question. dunno.

EDIT 2: I suppose it depends on your definition of “at events” in T11. At the day(s) of an event, or at the time, or at the place? <T11> At events, Teams may only produce FABRICATED ITEMS in the pit areas or provided machine shops, as defined in the Administrative Manual, Section 4.8, The Pit.

I know you are looking for instances of major changes to a machine throughout the season. But, one of the things that seperates the supposed “elite” teams from the rest of the pack is the willingness to not accept even minor issues with a machine.

At Champs, I watched teams come out match after match, having the same little issues of not being able to feed into their robot because of a poor hopper design…or not being able to pickup discs…or not using specific locations on the field to score from, or many other issues teams could have during a match.

I know that our team is not willing to continue to experience the same problem more than one time. As an example, when we miss fed a disc into the hopper, and it flipped into the shooter, we added a piece of lexan to block it from happening again. This didn’t happend very often, but we were not willing to accept this as a part of the design and just live with it.

These little fixes add up…they save time in a match, where even 1 second counts.

From a major addition standpoint, we were not able to finsh installing our climber during the build season. After we were eliminated at Waterford, we added the remaining peieces and strung the cabling for it. We continued to work on it through Troy and into MSC, before it really started working well.

The ambiguity of T11 bugs me. There needs to be a more comprehensive definition of what it means to be “at [an] event” than is provided and thus T11 can be interpreted in a couple of ways.

One interpretation might be that, while on-site at an event, you can only produce parts in the pit or in the machine shop. You could not, for example, have someone sit in the stands with a hack saw and chop a part in half. Another interpretation might be that, during the course of an event (e.g. from unbagging to rebagging), you may only manufacture parts AT the event.

The best thing you can do to allow room for change is to make sure you have 10+ pounds of weight to work with. I know this is difficult, but having some freedom weight-wise really let us keep improving our robot throughout the season. Our changes were minor; changing the flaps in our magazine from L-channel to C-channel, changing our intake drop ramp from sheet aluminum to carbon fiber, changing the mounting method of our shooter wheel to facilitate faster fixes, and adding a feeder station intake that we hardly ever used. All of these changes added a little weight but we bagged with a 105-pound robot. By the time we were at champs we were 115 pounds.

(Warning: Long Post)

Leeland,

I’ve been doing FIRST for 11 years now and I must say that this year will always be remembered. I started out with 1126 in their rookie year and moved on to mentor other teams like 229, 1930, 2228, and 3181. I have only come back to 1126 these past two years. When I was on 1930 and 3181 I was either the lead mechanical person or was one of them. Now back to why I will remember this year; I only missed about 3-5 days total (mostly Sundays) from kick-off until we played at FLR. It was the most hours I have ever put into a team and I worked my butt off, so did the students and a few other mentors. The reason was that we were not done and I was not satisfied with the robot yet.

I was part of the climbing team. We had a design VERY similar to 118’s corner climber. Wheel, cog and ski pole idea. We scrapped it week 4 (saturday morning integrations meeting). By the end of that same day we already had a 10 point passive 1 second climber designed and went through a few iterations. It took a few more days and meeting to button up all the small things we bypassed with the quick design and build.

The shooter/elevator team was struggling with their designs. I was coming in on the voluntary meeting days to provide support for them. I did not want to see the team fail and I was going to be there every second to make sure it didn’t happen.

As you know of our past we always strive to improve upon our designs.

Luckily for us this year, we had a whole month gap between FLR and Buckeye.

We competed with an alright robot at FLR. We were not able to get our shooter/elevator/floor pickup to work at all together. We worked until the last hour within the rules and brought in an exact 29.8 lbs of parts into FLR. We ziptied the elevator cables to the top and only acquired from the human player.

We did somethings different in the first few meetings of the year where we picked our strategy. The strategy team broke up the items that wanted to achieve into levels.

Level 1 was to be able to acquire from human load, level 2 was floor load. Sadly, our floor load was developed much earlier then the human load. This was only due to the human load being connected to the shooter assembly which was developed very late in the season.

The floor pickup was truly the best part of the robot. We actually had the floor pickup in the robot the entire time, just was ziptied up and out of the way because we couldn’t use it yet.

As you noticed, we had a sub-par shooter at FLR with a so-so mechanical build of a shooter and software issues until saturday morning. There was continuous improvement and finishing throughout FLR. We got away with smart strategy and quick changes to the robot and a bit of schmoozing to make our way into the eliminations where we found our real calling. Massive Defense. Going away from FLR with a Silver really made it all align into perspective. We all knew the robot had potential, but we were always missing the little keys to make it happen. We saw glimpses of the potential it had and I wanted a gold.

In between FLR and Buckeye was a whole month and an intense extension of build season really.

Once again at the end of FLR, we took off the shooter/elevator and brought it back to our shop. We used one of our old bases and mocked up the robot with our spare floor pickup and our real shooter.

In between FLR and Buckeye we designed a new shooter system and 2 different elevator systems. We also upgraded our floor pickup and a few other things on the shooter/elevator system. Once we had the floor pickup working out too, we also had our programmers develop a 5 disc auto. Sadly, we ran out of time and were only to be able to get the elevator system to only be able to pick up 2 discs at a time. This once again changed our strategy.

All these changes were done well because of the amount of CAD done. Because we all know hands off once the robot is in the bag, we must trust that the CAD is 100% and go with it. We ran into a few problems when we needed to move the entire shooter over 1/2" in the robot and cost us about an hour in the pits one of the thursdays.

In the end, it is hard to foresee problem areas in the short span of build and even harder to see them during competitions. The usage of CAD has really helped in the improvement between regionals.

One of my favorite teams that I always look up to is 67. Adam’s posts is spot on and the reason why they are successful on the field.

Good enough is never good enough.

If there are any other questions or ideas, I’ll be happy to post/answer them.

Hope you had a blast and learned while in STL. IT was hard to sit home and watch it through the computer. I will try even harder in the upcoming seasons to not have to feel that pain again.