DEWBOT X Design Philosophy

From DEW Robotics
Jump to: navigation, search

More so than in any past year, the team pursued perfection in executing DEWBOT X's design and construction. This can be best seen in the attention to detail, the modularity and serviceability of the robot. It is also visible in the repeated renditions of some parts, which we made over until just right.

Our inspiration

During 2013's Einstein Semifinal 2-2, alliance captain 303's (Test Team) battery cable failed, leaving us one robot down for the match. This is not a critique of 303, but of ourselves, for this failure could have been ours, had fortune been different. 3476 (Code Orange) and ourselves put up a determined fight, but lost this match 200-225. While it is impossible to say that we would have won (and moved on to finals) had this not happened, it made it absolutely clear that if we have the ambition to continue to play at this level of competition, then any failure could become (competitively) fatal and that there was no detail too small for attention.
This further reminded me of a 2012 visit the team made to sponsor UTC Sikorsky's Coatesville plant. I was very impressed by the attention to the processes which assured quality of the helicopters produced there. Of course, with any aircraft, quality is literally a matter of life and death. So impressed as I was, I did not immediately see how this focus on quality applied to FRC 1640. After all, we just make and compete with robots! No-one dies if our robot breaks.
Our 2013 Einstein experience, together with the team's hard goal to Go back to Einstein in 2014 and do it right crystalized the acute need for quality in our "product". Now that we finally get this, I hope we don't forget it.

4 Bolts

Quality is a tough thing to sell to Judges, and we did not do a good job of this. Quality doesn't look sexy, or technically intricate. It doesn't even look difficult. So how do you pitch quality?
Our pitch was 4 bolts. We could remove and replace any major mechanical system on our robot by removing and reconnecting 4 bolts. Why 4? Because it is a reasonable number for secure attachment for any small-scale (FRC scale) system. It shows that we've thought through service requirements and reduced service work to a minimum. It indicates that spares are either modular systems, or at worst ready spares.
Not a bad pitch. Didn't work, though, so it cannot be called a good one either.
What did work, however, was quality. Following our spectacular shooter failure at Hatboro-Horsham, DEWBOT X was the robot that very rarely broke, and when it did, was rapidly repaired. Our focus on quality resulted in reliability improving steadily throughout the competition season.

Modularity

Our pivots have been modular since 2011. This year, though, we went all-out for modularity to facilitate serviceability:
  • Pivots: consolidated Anderson power connectors and PWM sensor connectors make for faster pivot changes (not needed, because no pivots failed during the competition season)
  • Slingshot Elastics: originally attached directly through eyebolts, this was changed to threaded chain link connection to speed replacement time (not a common task, in practice)
  • Winch: modular like pivots. Anderson power connectors. Winch replacement was a common task.
  • Roller Frame: also modular. Anderson power connectors. Never actually replaced. Materials selection was critical to this system's reliability and success.
  • Rivet Nuts: By-in-large, systems were connected to the chassis via permanent threaded connections provided by rivet nuts. These greatly facilitated system changes by eliminating the need to place wrenches on back nuts, and positioning back nuts for reconnection. We started using rivet nuts in 2013 and expanded the practice enormously this year. Requires good design before building.

Ready Spares

Spare parts in 2014 were ready to use. Spare cylinders included the appropriate pneumatic connections already in place. When designs changed and called for different connections, connections on spares were changed as well.
We still brought a lot of COTS to competitions: Bolts, nuts, washers, stock Al & polymers. But the expectation was that if we have a motor, actuator or sensor, it is in a form ready for use. Sensors have soldered connections and PWMs. We brought no loose CIMS, BaneBots or AndyMark motors: we brought Pivot, Winch and Roller Frame modules containing these motors.
Spare winch straps were cut to length.

Get it right the first time...

The Drive Train was right from the start, but since it was closely derived from DEWBOT IX's, this is perhaps not such a great accomplishment. Still, there were no mechanical issues with the pivots through the FRC competition season other than tread wear (wheels were replaced at Championships for this reason).
The chassis was well thought out. Mounting points for pivots, bumpers, shooter and roller frame were all provided via 1/4"-20 rivet nut permanently secured to the chassis. These rivet nuts greatly improved serviceability by eliminating the need to get wrenches onto back nuts in the tight quarters of the chassis. Channels welded into the chassis also engaged fittings on the lower parts of the bumpers, securely locking the bumpers into place. Gussets provided rigidity and strength. No welds broke during the FRC competition season. The chassis design followed DEWBOT VII's, but with extensive design changes. The design frame perimeter of 111in provided no issues with inspection.
Likewise the shooter frame weldment was well designed and executed. No broken welds. Careful CAD motion studies identified a potential interference between the shooter frame and the pivots, allowing the design to be changed to avoid this problem before any metal was cut. All shooter frame elements were finished on the mill to assure accuracy.
Most major electrical components were mounted on the carbon fiber electrical panel. The panel itself was a structural element of the chassis. The panel was easily removable for service. Components were arranged logically and neatly. This year, the PDU was mounted on spacers, which allowed wires from the PDU to run underneath, facilitating wire organization. Connections to all motors were via 40 amp Anderson connectors.
Bumpers were no afterthought this year. Each bumper was designed with (4) points of connection to the chassis frame: (2) high bolted connections (like those on DEWBOT IX), and (2) low-mounted fittings which interlocked with welded channels on the chassis frame. Unlike DEWBOT IX's bumpers, all of the bumper mounting hardware was through-bolted with 1/16" steel strips on the back side providing assurance that these bolts could not pull through under load. These bumpers just didn't move once secured. Cloth covers ripped (and were quickly replaced), but the basic bumpers survived this very aggressive game without damage. Thumb bolts were used for bumper connection.

...but if you don't, do it over until it is right

  • The polycarbonate side and rear panels were remade from scratch three times. In each case, the protection provided to the robot's interior mechanisms was improved and extended.
  • We originally anchored the roller frame to the chassis using a pair of purchased bearing block, but quickly replaced these with CNC 6061 Al blocks when the originals broke in practice.
But the greatest example of improvement over time was with autonomous. At Hatboro Horsham, our autonomous barely worked except for mobility (5 pts). In the first 9 matches (including practice match), we attempted autonomous scoring 8 times and succeeded once. No team video after this, but I recall we largely gave up trying.


We blew it, then recovered

We experienced a nearly complete failure of the Shooter system at Hatboro-Horsham, our first competition. The Shooter failed to fire on most occassions, relegating DEWBOT to a ground passing and low goal scoring role. This was a surprise because this system had, we believed, been thoroughly tested.
It turned out that the shooter release failed due to insufficient air pressure to release the winch's dog gear. Testing at the CCIU had been flawed in that we did not operate the Roller Frame actuation in practice as often as we did in competition and this Roller Frame actuation was the main air user. In addition, air use calculations (such as we had done for DEWBOT IX) had not been performed for DEWBOT X prior to Hatboro-Horsham. Frankly, DEWBOT X had not been identified as having large compressed air requirements and as a result, the compressed air system (VAIR 0090 compressor and (2) 35 in3 storage tanks) were thought to be sufficient. Not nearly. Leaks in the pneumatic system exacerbated the problem.
Following Hatboro-Horsham, the air consumption analysis that ought have been done earlier revealed that we needed a higher-capacity Thomas 405ADC38_12 compressor with (7) 35 in3 storage tanks. Video from Hatboro-Horsham helped to clarify the actual use patterns of the pneumatic actuators. We tested compressors to determine the capacities. Significant air leaks were also repaired.
The team did a lot of very good and thorough work in investigating and correcting the shooter failure. Of course, this should have been done earlier.

The Payoff

In 2014, DEWBOT was the robot which just (almost) never quit.