Quote:
Originally Posted by Lil' Lavery
@Racer26
Several of the assumptions you make in your post, which parallel some of the decisions made by the Ri3D/BB teams, are not entirely true. While there are a certain degree of truth to them, to see what these builds did and use them as supporting logic for those assumptions is risky.
For instance, "acquire and pass" are not the same thing, nor does either equate directly to being able to score in the one point goal. There are plenty of robots with ground acquisition systems that stow the ball in such a way it is not trivial to pass using the same mechanism. Most of the Ri3D/Build Blitz teams are examples of this, as their eventual storage/launching mechanisms do not permit a clean release of the ball back to the intake mechanism. One can make similar arguments about several of your other points.
More so, you're viewing the conversation backwards. You're using the finished design as evidence of the "optimal strategy," when in reality strategy should drive design. To again use Karthik as an example, look at the priority list that the Build Blitz team he participated in came up with. Team Copioli's final product doesn't really reflect that list particularly well, as its lowest prority (high goal) is more emphasized than a few of the higher priority items (catching and receiving from the human player). Beyond that, as Karthik has repeatedly pointed out during his presentations, being able to accomplish a task is not as valuable as being able to accomplish a task well. A lot of these 3 day bots ended up being closer to machines that are "5/10" at several different functions, rather than "10/10" at only a few.
I understand the value in showing methods to complete several different game functions, rather than focusing on one or two, during a quick build project. I don't even necessarily disagree with it. But don't attempt to use that to justify their strategy decisions for a team with limited resources. |
I could agree that an acquiring mechanism does not necessarily allow for passing or low goal-ing, but show me a robot that can pass, but not score in the low goal, and I'll eat my hat.
I stand by my view that acquiring and passing a BALL ought to be priority number two, behind drive, for the reasons stated (also, you can't score a ball if you can't acquire it first). Its certainly possible to build an intake that DOES allow for easy passing. I know 4343's intake does that, along with dozens of others I've now seen pictures/video of. (and referencing Karthik's Effective FIRST Strategies seminars? Multi-purpose mechanisms that are simple and effective are a Good Thing.)
I can certainly see how TeamJVN's ball positioning doesn't allow for easy passing by reversing the intake. I'm pretty sure TeamCopioli's would do low goal/pass via reversing the intake action though, despite them not showing it in the video. BoomDone's El Toro's folded in touch the ball, and should be able to push a ball back out lightly. O-RYON can reverse its wheels. Ri3D 1.0 should be able to push a ball back out of its intake.
Any robot that can pass a ball out over its bumper, ought to be able to score in the low goal. I don't know that TeamCopioli could do this, but they can probably drive up to a low goal, spew it out onto the floor between them and the goal, and shove it in. A box bot can shove a ball into the low goal just due to the bumper geometry. I'll concede that any robot might not be able to do it well, but I would guess that TeamCopioli could do a decent job of it.
I think what really happened, though, is that they discovered when they started building and prototyping, just how easy scoring high is. TeamCopioli prioritized it that low, simply because they knew that scoring in the low goal has 70% of the scoring potential and thought it was significantly easier to do.
Shooting high? There is a large sweet spot (which Aren found on TeamJVN's build, but most teams would eventually figure out that there exists an optimum shot parabola, IMO), so shot location accuracy isn't terribly important. Shot alignment isn't terribly important, owing to the goal being the whole width of the field essentially, and getting a consistent shot trajectory from your shooter seems to be a fairly easy task this year. Those three things put together make it reasonably easy to get a shot that is REALLY forgiving on where you shoot it from and what direction you aim it in, which has the added bonus of making it difficult to defend against.
I'll agree that the 72 hour bots all collectively are closer to 5/10 jacks of all trades than 10/10 one task specialists, but I suspect that's probably a good thing in terms of their effect on design diversity. If a 72 hour build showed off a 10/10 anything, it starts to give away too much of the challenge.
I also stand by my assessment that the 5 72-hour builds that were completed (along with scores of 6-week bots completed) all converge on a similar strategy in their design, because that strategy they converge on is close to being the optimal strategy. Why else would they converge to that strategy?