Talk:DEWBOT VI Bridgewater Battle

From DEW Robotics
Jump to: navigation, search

Necessary Repairs & Maintenance & Drive-Train

There are some necessary repairs needed. Also, some drive-train fixes & upgrades. I include these together. Suggest we:

  • Pull all pivots. Add spacers between driven sprocket lower 1" bearing. Check condition and refurbish as needed. All treads will need to be changed (they're beat). I will order more treads.
  • Remount magnetic absolute encoders to provide better protection to the connectors (e.g., rotate towards centerline). use wire-ties on these and other electrical connections.
  • Replace Steering Jaguars with Victors and test.

We should do this Wednesday, 2-June.

On Wednesday, 2-June, we followed the above plan, without finishing.
  • Pivots removed
  • (4) 0.520" PVC Spacers cut from 1" PVC pipe (we've got miles) - thanks, Douglas & John
  • General cleaning & tread replacement. Steering sprockets pulled from (2) pivots - other (2) will require pullers (Clem McKown to obtain)
  • Drilled holes for Jaguar > Victor switch.
Generally, pivots are in good condition excluding tread wear and carpet fiber accumulation on the transfer axle. No "broken" pivots as at PARC. All steering sprockets are 0.52" above lower bearing (same as model dimension - 0.521") except Pivot #3 at 0.33" (thereby our steering chain problem). Pivot #3 steering sprocket set screws were loose (not quite finger tight).
The Pivot Replacement Process was documented.
On Wedesday, 16-June, we continued, again without finishing.
  • Molly & Garrison joined us - welcome!
  • Steering Jaguars replaced with Victors. All drive Jaguars are black.
  • Pivots 1 & 4 reinstalled. 2 & 3 still in service.
  • PVC spacers installed.


Autonomous

Not effective yet. Main points seem to be:

  • Sensor for kicking does not seem to be reliable.
  • Angles for off-axis shooting need to be checked. They appear to shoot off-field is the robot is aligned to drive along the central ball column. I'll check, but we should test too.
  • Keeping in a straight line by blind dead reckoning is often ineffective.
  • Keeping the possessor off seems to work better (insofar as something which doesn't work can be said to work better).
My initial angle calculations for autonomous were incorrect, as observed at (BR)2. My bad. the real angles can be found (with math) in Auto Angles (xls format).
Without special (BR)2 rules, reasonable autonomous sequences are:
  • Autonomous Near Field - choice of either:
  1. Set ball C2-R2
  2. Set wheels to 30° CW
  3. Set robot to drive down C(olumn)2 (use tape guide on robot top for alignment)
  4. Stop after 1 kick (not critical)
or:
  1. Set ball C2-R2
  2. Set wheels straight with Calibration (delay is beneficial here)
  3. Set robot to drive through ball towards goal
  4. Stop after 1 kick (critical)
  • Autonomous Mid-Field:
  1. Set balls at C2-R3 & C2-R2
  2. Set wheels to 9.5° CW
  3. Set robot to drive down C2 (using a 2nd guide tape)
  • Autonomous Far-Field
  1. Set balls at C1-R3, C1-R2 & C1-R1
  2. Set wheels to 2.5° CW
  3. Set robot to drive down C1
Where not specified, wheels should be pre-calibrated, wheel angles pre-set, and kicker pre-armed by tether prior to the match. This requires planning with alliance partners before queuing up for the next match.--Clem 1-June-2010

Possessor

We've got a adequate possessor for demonstrations. In the heat of competition, it is ineffective. We tend to lose the ball or else overrun it. We really cannot effectively herd balls into the goal either. This is an area which needs work for IRI. Not sure the vacuum possessor is better, but let's discuss.

I'll say this for completeness, because I know we've beaten the single roller concept rather hard. But why is it that 1726/1346/2046/1676/1918/33(old 1730?/100?) manage it? They're not all great, but some really do work quite well. I mean, not JVN/Copioli-approved work, but not our "work" either. (Yes, I said 1676. At least according to Don, which is plenty good enough for me. And I think I do remember it...maybe not.) Or 230? Though that might just be magic. Not suggesting we get a bigger stick to beat the single roller horse; I'm just trying to get on your same page. Knowing what's actually wrong now may go a long way to producing a viable alternative solution, regardless of the platform. Though perhaps it's worth re-highlighting that part of the issue seems to be mimicking competition conditions at home. Might be nice to solve that before we start pulling the old design apart.
As for an actual alternative, could the 2669/1583* side wheel method be more straight forward? I honestly don't know whether the vacuum is a good or bad idea for our situation--there are certainly enough good ones out there, albeit quite a few bad ones too. On their face though at least, side wheels seem simpler (to me). (I take it we've discarded any sort of 2438 thing?)
*I don't think these are the best-known teams for it, maybe Foster can help me? I'm thinking of a little non-1583 n>4 (6? 8?) polygon bot that's really good at the side wheel method. You may have shown me the video. I'll look in the (later) morning.--Siri 18:46, 31 May 2010 (UTC)
I'm working on a lower roller (just below the kicker) which can be retracted for going over bumps. Really an evolution of the idea Mike Rizzo was pursuing. Retraction in this case would be purposeful rather than passive. The control to relax the kicker could also retract the lower roller. Arming the kicker would bring the lower roller back to possessing position. Kicking and automatic re-arm would not affect the lower roller position. Can't say yet how it will possess, but riding over balls ought to be impossible. I have a fair bit of this worked out on Inventor, but not everything yet. As always, the devil is in the detail.
Also, I've done enough conceptual work to understand that the installation of a Team 25-type vacuum possessor is probably feasible. This will, however, require more careful ball alignment (L-R) than we currently need.--Clem 1-June-2010
Cool! Lower bar is definitely best case, nice work.
Given its proverbial Satanic inhabitance, is it worth having other teams look at the common/uncommon alternatives? ...If there are people not doing something more pressing (granted that's approaching rather unlikely). I sense some fun to be had and lessons to be learned with side wheel alpha prototyping, if so. The top roller stuff would (hopefully) be educational and good from a completeness prospective, but I really have no estimate on productivity probability. To be honest, I'm not really sure what we've done thus far with all the iterations. (Floating? Pulsing? Killer Bees-ing?) --Siri 16:36, 1 June 2010 (UTC)
Since both lower roller and vacuum approaches seem feasible, it would be good to pursue both if we have the resources. I think that the possessor is quickly becoming our de-facto mechanical summer project. It's not a bad project.
If we pursue both, the robot becomes the test-bed for the lower roller. The proxy chassis would need to be called back into action for the vacuum development. We'll need to be careful not to execute anything in the proxy chassis which we cannot duplicate on the robot.
I am uncomfortable as a project manager not having a Plan B to cover points of uncertaintly. The performance of a low roller possessor is a clear uncertainty at this point. A vacuum possessor is a good Plan B. But if Plan A works, Plan B gets dropped (but we should not forget what we have learned).
By-the-by, running planned and controlled tests with identified independent and dependent variables and documenting (& analyzing) the results is the basis of knowing what we know--Clem 1-June-2010

Trirol Possessor

At PARC, Monty Madness and (BR)2, we were dissatisfied with our possessor's performance. While it seemed to perform well at the STEM Defined demonstration, in the heat of competition, it tends to either fail to possess the ball, or to cause the robot to drive over the ball.
We've observed that robots with effective possessors often have passive rollers at or below the ball centerpoint. Installing such a low roller on DEWBOT VI without a major redesign is not trivial. Key challenges are:
  1. Avoiding interference with the kicker
  2. Avoiding interference with the pivots
  3. Maintaining our ability to cross bumps without difficulty. A low-position roller intrinsically is in the way when climbing a bump
But not impossible. An articulated, low, 3rd possessor roller has been designed for the existing robot.
  • The roller articulates up and out of the way for bump crossing.
  • Two 3/4" diameter x 4" stroke pneumatic cylinders provide the actuation. Both run from a single solenoid. Restrictor valves will need to be installed on the solenoid vents.
  • A hard stop is provided for the possessor in the lowered, working position (the design does not rely on the pneumatic cylinders to provide this stop).
  • While rotating between the deployed and retracted states, the low possessor assembly does not break the frame perimeter.
  • It is proposed that the low possessor will automatically retract when the operator relaxes the kicker to cross a bump; and automatically deploy when the kicker is armed; normal kicking will not change the possessor state.
  • In addition, it is likely that the combination of reversing the driven possessor and retracting the lower possessor roller would be a useful, low-velocity means to positively eject balls into the goal from the goal's ramp (without having to drive the robot entirely into the goal).

View the detailed drawings.

Prototyping Suggestion

3rd roll prototype installed on DEWBOT VI
The third bar concept looks promising but we have tried several good approaches already without much success. Before building the proposed mechanism, I'd like to see us test the concept.
Testing shouldn't be hard. We can clamp a rod to the bottom of the robot where the proposed modification will sit. It doesn't have to be retractable, it simply needs to sit there. Slip a tube over it and it will be rotatable. We can then drive the robot around and see how well it will handle the ball before we try anything further.
There's no reason that we shouldn't be able to build and test this in an evening and it will let us know whether or not we should continue on this track or pursue an alternate path.
--Jon 09:39, 14 June 2010

30-June Update

3rd roll prototype was installed and very briefly tested. Key observations:
  • Possession performance appears unaffected by the 3rd roller - poor-to-mediocre
  • On the other hand, it is now impossible to override the ball
  • and herding now works very well, possessor on or off
Overall, some improvement, but not what we were hoping for. We'll test more on 2-July.

2-July Update

Initial test just added friction tape to the bottom roller. No observable improvement.
We were planning to replace the middle steel roller with fiberglass, but before installing the fiberglass rod, we tested the possessor without the middle roller. This gave a phenomenal improvement in possessing performance. This Hi/Lo 2-roller possessor really works! It hangs on to the ball. The Hi/Lo 2-roller possessor is definitely a pincher. The bottom roller needs to be high-friction and needs to not roll easily (or at all). The bottom roller also needs to be stiff (so steel is better than fiberglass). No ball overdriving problem. The possessor will lose the ball if the ball touches the Ball Dam Gussett on either side. Transverse driving is therefore a problem if taken too far, but it can also serve to center the ball. We tested filling in the top roll spiral gaps with friction tape, but this had no obvious effect. Existing spirals are definitely not beneficial (they do not center the ball), but don't seem to hurt performance either.

3-July Update

Given the complexity and fragility of the bottom roller pivoting mechanism, we started investigating vertical retraction. Prototypes with borrowed (with no intent to return) parts from an altruistic DEWBOT IV proved promising. This should indeed be a simpler and more robust method. It requires fewer (and less complex) parts and stress points, though it's slightly complicated under the possessor motor and gets rather cozy with the top roller (we've started shedding neoprene). The current system is also narrower than the original possessor: approximately the width of the kicker instead of the robot. However, given our inability and perhaps even disinclination (if the pincher versus centeror trade-off is as it appears) to center balls, this may actually be a good thing.
Additionally, benchmarking of other successful bottom rollers (which neither roll nor facilitate rolling) revealed an inordinate number of teams using 1/2" angle. Prototyping (with aluminum) seems to reveal why. The ball tends to roll (and thus fall out) much less than with even the misnamed bottom roller bar, and holds somewhat better in snake and much better in strafing. (Strafing is a beneficial aspect of our drivetrain could prove very helpful on defense against a certain looper bot at IRI.) The system isn't yet to death-grip standard, but I (Siri) wouldn't be overly surprised if changing the roller(s) surface(s) and/or speed (recommended at 2x robot speed) got us quite close.
Given the strength of our pincher (and more prolonged testing), we've begun to rather quickly eat away at our polycord drive belts. To mitigate this, we'd like to implement a holding torque or different duty cycle on the top roller when we have a ball in possession. This requires an accurate detection sensor, which is shaping up (slowly) to be a beam break. (We're running out of time for this to move slowly.) Some code has been written, though obviously not tested, for the sensor and initial prototyping appeared promising. Additionally, other teams have suggested different belt materials as needed.
We appear to be having sporadic DS-robot communication malfunctions, both during bootup and randomly mid-test. It'd be great to determine the root cause(s).

Mirror

Looks really good, but we should:

  • Verify that it stays within the frame perimeter.
  • It would be good to have a means of raising it during competition. A servo might be adequate.
  • A working possessor, or a sensor to kick automatically would be a benefit now that we can find the hidden balls.

3-July Update

  • The mirror was moved back 1" to keep it within the frame perimeter.
  • Initial prototyping and CAD took place for a servo raise/lower mechanism.
  • Possessor progress is looking way up (see also 2-July).