Cycling Optimization

#1

We’ve been asked in the past to put together tips for how to work on optimizing your robot for fast cycle times. Over the years 610 has built many robots that have been designed and almost solely focused on fast cycle times, our robots in 2013, 2016, 2017, 2018 are good examples of this. Most years it was a major advantage for us to design our robots around cycle times (‘13, ‘16, ‘17), but last year eventually it reduced our effectiveness as the game evolved away from us (‘18). As most teams probably realize this years game is going to be the fastest cycling game FIRST has ever seen, human loading and scoring on your side of the field should cut down cycles by a couple seconds and make this the fastest scoring game we’ve seen.

Throughout the years 610 as a team has developed a couple tips and tricks that we use to design our robot around and train our drive team to optimize cycle times:

Build your drive train for Acceleration not Top Speed

Most years we’ve made the decision to gear for a slower top speed and focus more on acceleration. We will plan out what our primary drive path will be (loading station to Pyramid in ‘13, low bar to tower ‘14), and optimize our gear ratios for that path. It usually ends up being a something in the 9-11 ft/s range. In ‘17 we built a 2 speed drive and it was helpful in an open field, but it felt like we had less agility around D than in previous years.

Simplify scoring alignment

We think a lot about how to make the scoring actions as fast as possible, this usually involved reducing the amount of time drivers take to line up the pickup and scoring of game pieces, and optimizing the mechanisms to score as fast as possible. Two of the best example of this are ‘13 and ‘17, in ‘17 in or quest to make the fastest gear bot we put a lot of our design effort towards placing gears as fast as possible. We included an active gear release mechanism, flexibility to slide horizontally, and eventually automated sensors to release the gear once the peg was through the gear and inside our robot. In ‘13 we simplified lining up to the goal with our pneumatic “bunny ears” to square ourselves using the first level of the pyramid. And throughout the season we iterated on the shooter to dramatically improve the speed we could shoot the batch of 4 frisbees.

Many teams use a driver and an operator. A key factor in simplifying scoring is to reduce the amount of synchronisation required by the drive team. In ‘17 for example, our operator only had control of the climbing mechanism and fuel mechanism. The driver had control of the gear mech, before it was automated. If you can’t automate scoring or pickup, often the driver having control can be faster than the driver having to sync with the operator.

Practice, you talking about Practice?

Probably the most important part of getting really fast cycle times is getting your drive team lots of practice behind the controls. But the most important thing we’ve learned is how to run driver practice. Usually the first couple days of driver practice are more about working out the issues with the robot, figuring out the optimal controls, breaking the robot and fixing it.

After most of the major issues are worked out with the robot we then start focusing on drills. We try and design drills that will represent the major functions of the game, in 2017 cycling to different parts of the airship, taking different paths to different pegs on the airship. Once the drivers get comfortable with those paths and cycles, we will start challenging with various tasks, like not dropping gears, to take perfect paths into the loading station so that they don’t have to re-align the robot, to not drive too far while decelerating into the airship. An example of how we would enforce this is putting pylons around the area around the airship to limit the space to decelerate and score the gear. The penalty of touching the cones, was a reset of the count and time, essentially to force to drivers into thinking about all the finesse needed to play a full 2 minute match.

Finally once your drive team has started to master the main actions of the game, this is where maybe the most important part of driver practice that most teams miss, Defence. We will usually have at least one of our old robots setup as a defence robot, and start challenging our drive team to drive through D. This is where they will start developing counter moves and learn the best ways to get around defence. The best way of dealing with defence is to never try and drive through it, but drive around it, and avoid T-bones. Our drivers will learn counter moves and fakes to get around D, once they start making the D look foolish you know you have a good drive team.

During competition our best tool for improving robot performance is reviewing match footage. We have team members who will record a full-field view of every match, then as quickly as possible download the footage and review it with the drive team. After every match our drive team and strategy team will review the match footage and have an open conversation about what worked and didn’t work in each match, focusing on what we can do to improve.

610 doesn’t build a practice robot, so we just finish our 1 robot as early as possible. We aim to spend weeks 5 and 6 of build, practicing then breaking the robot. Only to fix it and then continue to practice.

2017 Driver Practice Example 1

2017 Driver Practice Example 2

2017 Driver Practice with Defence

2018 Scale Cycling

Singular Focus

One of the biggest issues I see a lot of teams make is being distracted by other robots on the field and not knowing what to do next. A lot of time you will see robots get distracted and waste time struggling to decide what to do on the field and ending up doing being so unfocused during the match it hurts the alliance. Ensuring that every team on your alliance knows the game plan and sticks to it is very important. Our strategy and scouting teams work very hard before each and every match planning with our alliance partners the match strategy. The goal is so that every drive team knows where they are getting game pieces from, where they are going to score them, and how they are going to play D.

In most games where the actions of the other alliances doesn’t fundamentally change how you can score makes this easier to plan for, last year’s game ‘18 was really interesting strategically but also put a lot more pressure on drive coaches to react to the game and make decisions on the fly. This is where drive practice really pays off, if you pick the right drive team and train them well, for most matches your drive team should be able execute the plan independently from the drive coach. This will let your drive coach focus on making sure the whole alliance is executing the plan and reacting to how the match plays out.

Learn From Others

One of the best thing about the FIRST community is how open and eager everyone is to learn from each other. At the end of each build season we open our school and practice field to local teams to come by and practice, we’ve had some great times practicing with the likes of (188, 1241, 1285, 1310, 4476, 4946, 5036, ect.), and learned a lot about how other teams develop their robots. Most of the time the robots will initially be in rough shape, but having other robots on the field really changes your perspective on the game and increases the difficulty for your drive team. I learned my lesson with 1241 in 2016 to not judge a robot before their first competition, that robot challenged so many of my assumptions about the game and how many different ways you can be successful.

I’m excited to get back on the practice field in a couple weeks with our friends trying to improve and optimize our robots. Hopefully others can chime in and share their experiences optimizing cycle times, and how they focus their robot, strategy, and drive team for excellence.

91 Likes
Mid build season thoughts
#2

Seriously though, absolutely amazing post.

Adding to this, there’s a huge misnomer in FRC that distance is the determining factor in cycle times. While obviously a large component, with the speed/acceleration of modern FRC robots, the biggest area where time can be gained while cycling is during game acquisition and release. JNo’s post has some great tips on how to excel in these areas.

18 Likes
#3

Just shared this with my drive team. Also shared some videos from Einstein in 2013 for inspiration. That was my rookie year and first year at Champs. We had the opportunity to play with you guys in quals and subsequently watch you beat out what I thought were the best two alliances in the world (1114 + 118 in Semis and then arguably the best full court shooter in the world with 2169 in the finals… talk about running the gauntlet) with pure cycling efficiency.

This is the best thought out, and most valuable post I’ve seen on the new CD… which is a great change of pace. I’ll be sharing it with as many teams as I talk to.

1 Like
#4

Thank you for sharing your philosophy on game strategy and robot design. Your post explains perfectly why your team and our friends at 1477 were able to win Einstein in 2013 against two full-court shooters.

#5

With the number of motors allowed now (and NEO’s) you can gear your robot to be so fast it’s totally uncontrollable, one of the advantages of gearing lower is more controllability and less time spent readjusting your alignment. This year I think the main factors in cycle speed are going to be in order: time to align hatch / cargo while scoring, time to acquire hatches / cargo, driving to score, moving mechanism to correct height position. Personally I’m more interested in seeing unique hatch / cargo manipulators then the latest lift / arm that can get to position in 0.01 seconds :stuck_out_tongue:

7 Likes
#6

Excellent summary. As a smaller team, we realize we can’t do everything at the level of competitiveness that the top teams in our district do so we don’t try to compete with them. Instead, we try to be the most valuable first pick. Deep Space really plays to this as the more difficult scoring aspects are worth the same amount of points as low and mid-level options.

I would add one thing to this list and it relates to acceleration. We try to keep our center of gravity as low as possible. Many times, this makes elevators and the top scoring position not as valuable to us as to others.

And building less complex robots allows us more practice. We practice a lot but never enough. Taking video and breaking each cycle down into segments shows us where we can make up the most time.
We expect the slowest part of the cycle this year (as most) will be lining up to pick up a hatch panel or score a hatch panel. For that, we are going to compare what our driver can do manually vs what our programming can do once the robot is close to the target.

2 Likes
#7

Thanks for making this post. 610 consistently produces some of the most thoughtfully designed robots in FRC year after year that epitomize Karthik’s golden rules of building within your limits and specialization. I know personally at the start of every build season I find myself think “what would 610 do” quite often for inspiration.

I’m curious how you think robots who are capable of only scoring 1 game piece will fair in 2019. While I think this game offers a lot of interesting options for teams to specialize and make meaningful contributions to their alliance, I am somewhat skeptical as to how well an alliance of specialized teams would fair at the highest levels of play.

1 Like
#8

I think robots that can cycle only hatch panels to low goals can be very successful in this game, scoring hatch panels will be more difficult then scoring cargo for a couple reasons:

  • Only 2 HP stations vs 2 HP + ground pickup for Cargo.
  • The alignment accuracy needed to score Hatch panels is much greater than scoring Cargo. By this point almost every team that has done FRC should know how to manipulate balls, and the scoring area for cargo is more forgiving.
  • Alliances will want to score Hatch Panels first as they are required for scoring Cargo. Thus I see most teams starting Sandstorm with Hatch Panels loaded, potentially placing only null Hatch Panels, and doing their first couple cycles with Hatch Panels.
  • I predict triple cycling will win Champs again this year. :slight_smile:
1 Like
#9

Beautiful write-up, and something I think all new strategy mentors should read. Most of our mentors come from industries where longevity reigns supreme and cycle times are considered, but are far from the most important item. It takes some time to realign thinking to score as quickly as possible, even at the loss of robustness.

The other thing I think that gets looked over that wasn’t mentioned too much in here is acquisition. Many teams (103 most certainly being one of them) look far too much at the scoring of game pieces and neglect the importance of acquiring them. I’d argue that acquiring elements quickly is far more important than scoring them quickly because one almost always precedes the other.

What exactly are you considering “Triple cycling”? Scoring three game pieces in the sandstorm?

#10

This post is by far one of the best I have read this year on chief. This information is truly timeless, and will be by far one of the most beneficial things to teams that can embody it. Thank you!

#11

I believe that you are absolutely right about acquiring game pieces first. I have seen many teams in the past build elaborate shooter, with turrets and amazing vision processing, only to have a poor mechanism to move balls from the pickup to the shooting mech. It is certainly important to consider all aspects of “life” of the game piece in the robot from acquisition to scoring. On 610, we usually prioritize pickup of a game piece over the scoring, often using the language “touch-and-go”, which means that ideally we want to have control of the element once it contacts the robot.

That being said, another factor to this is understanding the zones of gameplay, this year for example you are protected (more than when scoring) when you are getting a hatch panel or cargo. You have no protection when scoring. It is an interesting debate to talk about which one to optimize. Protected zones can be a factor in determining where optimization will be more beneficial.

By triple cycling he means 3 robots just spending all match cycling no dedicated defence robot. In essence taking 3 mid tier teams with high accuracy and having them essentially get as close to a perfect score as possible.

5 Likes
#12

All three robots on the championship alliance will be primarily scoring game pieces all match and defence played will be limited, see world champs in '13 / '16 / '17.

2 Likes
#13

Couldn’t agree more, there is so many options for scoring it will be hard to play strong defense against world class robots, even more so than 2017.

But we arnt practicing for Einstein just yet so it will be important to prepare for the army of 3rd bots who will be sent across the field this year.

1 Like
#14

Awesome stuff here! One of my favorite posts on CD in awhile. 610’s 2013 robot is still one of my go-to case studies for my Strategic Analysis presentations.

The top 2 things I preach in Cyclic Games are

  1. Thoughtless game-piece acquisition. It all starts with Touch it - Own it intakes. Whether this be ground intake or HP intake, making sure that there are no slip-ups on the game-piece acquisition is CRUCIAL

  2. When the Game Piece enters your robot, it goes to the same place, in the same way EVERY TIME. This one is going to be huge this year, and the #1 reason why I am ignoring the Velcro (Hook & Loop Tape :roll_eyes:) on the hatches. Handing the drivers another variable to deal with for every cycle is the exact opposite of what you want to do. It makes lining up & scoring the game piece that much harder every cycle. Eliminate all unnecessary variables.

As far as gearing Drivetrains goes, the sprint distance criteria is really the only determining factor I use when designing transmissions. My sweet spot is right around 12-13fps. If anyone is having trouble grasping this, do not worry! There is a drivetrain calculator made by JesseK that does all the math for you! Play around with some gearing options & cycle distances until you find a Sprint time that you feel comfortable with.

8 Likes
#15

Just as of recent my has been adopting this strategy of cycle times and limiting the cycle times throughout the season. I love the detailed information that you guys have given in this thread. This will affect more because of sandstorm and lower resource teams not programming auton. Our team at kickoff calculates an average cycle times for certain activities and we start structuring around that, doing that the last two years was key to our success

#16

I completely agree with what you have to say about using the Touch it - Own it strategy for game piece intakes this year. Another point to note is the release of the game piece. Like Steamworks where many teams were able to push the gear onto the peg with pneumatics (thus speeding up cycle times), release of the game pieces this year should be fast and account for error on the driver’s part (that is to say, you can still score a hatch if you’re not 100% square with the face of the cargo ship or rocket).

#20

com-video-to-gif

6 Likes
#21

Good writing. Would you consider doing conferences, there will be may teams who would learn a lot from this. Understanding gear ratios, free speed and actual speed will help. Again depending on the field, you may never reach the full speed, but having good acceleration definitely helps.