Monday, February 6, 2017

DDT: (Onyx) One filament to rule them all!

Upgrades: v3.5

DDT, my antweight (1 pound) combat robot, has seen much success and acclaim due in part to its adoption of the Markforged composite FFF technology. It has lead the curve in additive manufactured combat robots and at the time of this article holds the #2 ranking for its weight class, worldwide. Markforged has done a few articles about DDT and you may have seen a DDT replica as a display item at trade shows across the US.

Failed pulley at FI 2016

As wondrous as composite 3D printing is, materials cannot compensate for design or fatigue. DDT suffered from some belt/pulley issues at Franklin Institute 2016 so I elected to update the weapon drive to 1/4 inverted MXL belts just like Silent Spring's weapon drive. This of course meant I had to rebuild the frame to accommodate for the thicker weapon pulleys. I swapped to Maxon 13mm gearheads while I was at it; this would save me space and weight for the heavier transmission system.

That DDT frame was also a bit behind my 3D printed iteration curve. It still featured fiber in every layer and Silent Spring proved this was not necessary. But hey, why not take one step further? Fibers significantly improved the stiffness of frames made of nylon but the release of the new Onyx material, nylon with a percentage of chopped carbon fibers, proved that stiffness and strength could be achieved with that material alone. I decided as a challenge that the latest version of DDT would use Onyx material only. If this is successful, builders can create awesome parts for significantly less cost! What I am referring to is the new Onyx One series of printers that were released just a few months ago. They are limited in material capability compared to the Mark Two but you have access to Onyx prints for $3-4k which is approximately $10k less than the Mark Two. That is something hobbyists can more readily afford!

As a test for the material, I elected to change as little of the design as I could. That should hopefully draw a direct comparison between the materials since the design elements stayed the same. An obvious variable is that DDT may not fight the same opponents at the MASS Destruction combat event, but we will have to assume that the forces the robot experiences will be similar between opponents. For the curious, that's a 50% infill. If you go fully dense in a large part you aren't fully utilizing the power of additive manufacturing!

Eiger slicing shows only Onyx!

With the continued software updates, the Markforged printers continue to impress. The support structures easily peel away and leave me with a usable part from the printer in less than 5 minutes. I had to bore out the motor mounting holes because the support structure droops at the top of the circle but I could have easily compensated for that by designing it as a teardrop shape. Remember, Onyx material can print at ~30 degrees from the horizontal without needing support!

Fresh print right off the Markforged

The electronics transplant was quick, even with the motor swap. It looks like when the disk cracked, part of it flexed and actually hit the pulley which caused the delamination. a portion of that energy actually bent the stator mount in my Hacker A20L! No wonder the belt distances were too close!

The blade mount was updated for the double washer load distribution assembly which I found absolutely necessary to keep everything stiff in plastic insect robots. To summarize, the large face aluminum parts increase the amount of area subjected to forces and is less likely to plastically deform after frequent shock loads. Small but precision spacers keep the inner races in compression while the outer races can float axially; this keeps things from being over constrained which allows longer bearing life and higher efficiency.

The wheel guards are 1/16" nylon machined on the waterjet cutter. They were brushed with scotch brite in the long direction to match the layer striations and matte finish of the Onyx material.

The last part to make was the blade. I still had the 'bash blade' from old versions of DDT but I needed a new standard blade for use against most robots. I managed to cut one last blade from the 3/32" O1 plate that I had but this time decided to temper the blades. At first quench I achieved full hardness at 65 RC but them tempered the blades down to 54. This should give me more toughness but plenty of hardness.

The completed robot weighed in at 15.5 oz which is lighter than previous versions of DDT. So naturally, I spent some time in decorating the robot with vinyl stickers.


MASS Destruction 6: Watch the City Wreckage!

I took this new version to our local competition, MASS Destruction, which was hosted by the Charles River Museum of Industry and Innovation and ironically sponsored by Markforged. There was a great turnout and many fearsome robots including the defending champion "Foiled", a undercutter robot which featured a turbine-spoke disk to give it additional down force; this robot was printed on a Mark Two printer.

MASSDestruction 5 champion, "Foiled". Photo by Alex Crease

Long story short, the pure-Onyx frame worked wonderfully! The robot took and delivered hits without any unreasonable failures in the frame. As expected, when the frame is hit, it will take damage in that location but that is expected of any material. If anything, the overall damage was minimized because the hollow structure simply crushed in a localized area. If there were continuous fibers, the damage may have propagated to nearby sections of the chassis.

Damage from Foiled in our first match at MASS Destruction

DDT went undefeated through the day until it suffered a heartbreaking electrical problem in the last moments of the 'winner takes all' finals. That landed DDT with a 4-1 record and 2nd place. Who won 1st? None other than "Foiled" so really we can't be disappointed here. However, in my opinion, the real winner are the Onyx One printers because we've just shown that Onyx alone can hang in with the best of insect-class combat robots!

DDT takes 2nd place at MASS Destruction 6!

Fight Breakdown

W vs Foiled
Right off the bat, I get the fight the champ. Foiled has more MOI than DDT's blade and its made from a tougher material (AR500 vs O1 tool steel). His wide stance and airfoil would likely mean he would look better if we exchanged impacts so my first thought would be to attack the wheels. That sounded like a great strategy but if I missed he would likely take my wheels! With the faster and more nimble drive, I decided that I would slowly edge his blade with mine at full speed and try to machine his blade down. What ended up happening is that my blade slid under his and started clipping his bolts! That caused him to go unbalanced and vibrate. Then I decided to be more decisive and attack his wheels to some success but at the same time taking some damage (as shown above). The match went full distance and the judges awarded DDT the victory.

W vs Micro Mystery
These girls were newcomers but it looked like they had experience with combat robotics before. The robot was no slouch; it had a very long undercutting bar which by eye could over reach DDT. I hoped we might be the same blade height otherwise it would end poorly for me (Silent Spring v Circuit Breaker @ Bot Blast 2016). But when they loaded in, Fred flipped over their robot. I think they had delayed which subjected them to the 'wheel of misfortune' in which one of the punishments is starting inverted. Fortunately for me, DDT's sleek profile was able to sneak under their overhead blade and hit the chassis before their blade could hit mine. After a few hits, the battery became exposed and by event policy, their robot would lose by KO.

W vs Eclair and Bonbon
Ian finally upgraded from the toy transmitter and so he had full control of his bot. When he has full control, hes quite dangerous; his beetleweight "Quicksand" is evidence to his tenacity and driving skill. My strategy in this match was to maneuver around long enough to get to speed and deliver light, jousting hits. After removing some cosmetic pieces, I tried the 'spin slice' maneuver to hit higher up on the bot. One hit flipped him over and disabled his power. Tap out victory for DDT.

W vs Eclair and Bonbon
I aimed for the same strategy as the match before but this time we saw a more exciting result. DDT caught a lucky grab on part of Ian's bot and an explosion of parts signaled a KO victory.

L vs Foiled
Now in the finals, we find a familiar opponent! Foiled reversed his screws fastening the blade such that the button heads were facing down in an attempt to try and keep them from shearing off. But looking at our blade marks from last match I knew there was still a good chance of the same thing happening. My updated strategy was to disrupt his drive by forcing the blade to go unstable; I was able to slowly pull it off and started going for the wheels. As Foiled was no longer able to be the aggressor and took more damage, I thought I had the match won. However, one last tap dislodged a main power connector that was damaged in a previous competition and DDT shut down. Alex was a great sport and tried to revive DDT but to no avail; DDT was counted out in the final seconds of the match!

Thursday, January 7, 2016

Hasty Upgrades and MASSdestruction @ the Artisans Asylum 12/05/2015

Since Franklin, I've only been searching for more ways to use/abuse the MarkForged composite 3D printer. DDT has shown great promise of the device in combat, allowing high design flexibility and better material properties than anything I could have machined (not an advertisement I swear!). Next, I plan on applying this technology to my other robots, starting with Silent Spring; I anticipate printing a new frame/design for the Dec 5th MASSdestruction event.

I also apologize in advance because I have recently seen the new Star Wars twice recently in addition to all 6 movies prior so this post might feel more Force'd than usual.

A New Design

Converting the v1 Silent Spring to v2 should not have been very difficult since DDT v3 was so successful; there were little to no weaknesses that needed to be addressed. As a result, I followed the same methodology in designing DDT. I'll elaborate more on my design methods in a later post (I have most of it written up I promise!).

As a result, the v2 Silent Spring looks unsurprisingly similar to DDT. I at one point made an argument against myself to remove the pockets but in actuality they have an added benefit to the MF machine. The pockets direct fiber paths in slicing which in effect creates higher density of fiber in places deemed more valuable.

One place the designs may have differed is at the wheel output shafts. On DDT, the wheels are directly drive off the outputs of the spark gearmotors but a similar scheme may not be as effective here because the drive motor options for Beetleweights have shafts which are too small. For example, the B-series motors and Kitbots '1000 RPM' motors  have shafts 4mm in diameter which is disproportionate to the 3mm shafts used in Antweights, like DDT. Therefore I had whipped up a solution where the gearmotor shaft would couple to a larger diameter shaft which was further supported by additional bearing sets.

I ended up ditching the idea when I learned that Mark at ROVParts was selling B-series gearmotors with hardened S7 output shafts. Simplicity woooo! The new design intends on using wheel guards anyway so I HOPE this will be a moot point as such on DDT.

Although the robot was designed as a unibody, the printer volume actually could not handle the Beetle-sized DDT frame and as a result, the frame corners had to be chopped into four additional parts which would be combined into the unibody later using screws and adhesives.

The internals, like DDT, would require some minor electrical work. I purchased a new motor (NTM 35-36) and new ESC (Turnigy Plush 40A) for this project. To make everything fit, I would need to move the capacitors off the board and into a floating unit as pictured above. If you do this, be sure to keep the wires short and thick as they are bus capacitors. Im not entirely happy with this arrangement mostly due in part to the capacitors off the BotBitz ESCs. It is possible I shouldnt need those since i have bus capacitors from the weapon ESC. Instead, I could combine the capacitors and use electrolytics of equivalent capacitance.

The Printer Strikes Back

Just like the new DDT frame, this frame proved to be rather ambitious for the printer. This is by no means saying the printer is incapable, but rather the design needs to be modified to print more reliably. The access hole for the electronics is a removable panel on the back. This opens up to a cavity in the frame which is created by printing the roof over a bed of support material. However I discovered that the printer could not reliably make these features as so the rate of failure was fairly high. As a result, I managed to print the extra items such as the corners, back plate, and wheel guards but not the main frame itself. I will need to develop a interim solution for the Dec 5th event.

Return of the Old Design

With only ~2 days left, I began a mad rush to retrofit the v1 frame for the competition. As it stood, Silent Spring could compete with the damage it took at Franklin. However, some upgrades were planned regardless and I wanted another chance to try the MarkForged printed material. At minimum, the electrical system was better organized with shorter wire lengths. I also replaced the Turnigy AE 30S ESC with a Spider ZTW 40A ESC. Hopefully this will alleviate spinup and overtemp issues experienced in combat. The Spider ESC was also preloaded with SimonK firmware which has been making quite the buzz lately for its excelling sensorless startup performance; between the higher current and SimonK, I had high expectations for this new ESC.

The v1 frame files were modified lightly to better use additive manufacturing. The back plate was extended, thickened, and pocketed to provide coverage over the motor pulleys and provide additional space to hide the BotBitz ESCs.

In the A-frame section, the trusses were thickened and the motor mount holes were made slightly taller. These changed allowed for more carbon but also reduced the overall part count (the aluminum "C" spacer could be removed). Another fun change was switching the round hole to a hex hole in the interface wings. Normally, this section required a large hole for a socket to reach the locknut. Now, the hole could be changed into a hex designed to fit the 3/8" hex such that no additional tooling was needed to tighten the outboard shoulder screws.

Both of these printed parts finished at 1pm Friday, just hours before the event Saturday morning. Both parts came in lighter by ~50% compared to their SolidWorks projections for "solid" material yet were about as a stiff and a 7075 aluminum frame.

I took the opportunity to make a few other changes as well. I fell in love with the Fingertech Robotics snap hubs so I decided to make my own. I further integrated them into the drive pulleys as shown above. This provided a minimal solution to indirect drive with easily swappable wheels.

A few hours later, the electronics were transplanted into the new frame to yield the final result.

Currently the downside to printing with these composites is the inability to effectively weight budget. When I print these parts, I don't yet have a way to estimate the total mass of the parts and therefore I often overestimate to guarantee all parts comply with the weight class. This means I often have leftover weight which goes unused.

To attempt to top off the scales before the event, I printed some press-on shaft covers to prevent vertical spinners from dinging the ends and bending the shaft. They worked about as well as you might imagine for a last minute addition (hint: total crap).

MASSdestruction @ AA, Dec 5th, 2015

Thanks to Will Bales and the food I bribed him with, I was able to get a ride to the Artisan's Asylum in the wee hours of the morning. We unloaded and shared a pit table tucked behind someone's chair making bench. This time the attendees were looking pretty serious. Among them were Kyle Singer (of the Ripto robot series) who brought Dick Dastardly and Brandon Nichols who bring his multi-time champion Mondo Bizarro. Best Korea was now sporting an enormous 10" undercutting bar which out ranges pretty much every Beetle in the world. I also think the number of WPI kids in attendance doubled since last time. Absent from the fray this time was Jim Fox and his kids who usually bring Puppy, a durable and powerful wedge robot that beat DDT twice last MASS-D.

Both robots did well; DDT went undefeated and took 1st in the Antweights while Silent Spring suffered only one loss in the finals and took 2nd.

Silent Spring (5-1, 2nd place)

I felt pretty good about the new frame and new components in the v1-style frame; based on its previous competitions, I expected about the same in performance and it did not disappoint. However, the flaws noted previously in the v1 style frame also manifested themselves which ultimately lead to its demise. This event demonstrated the importance of the v2 modifications in any future iterations of Silent Spring.

  • Win v Dick Dastardly
  • Win v Best korea
  • Win v Herpaderp
  • Win v Dembitz
  • Win v Ginger Baker
  • Loss v Mondo Bizarro

Silent Spring got off easy in many matches. In the first match against Dick Dastardly, Kyle's radio suddenly went bonkers after the first hit and he lost control of the robot; Thus losing by disqualification. Best Korea definitely boasted the superior range and even sliced a third off one of SS's tires but his reign of terror ended early when his belt came off. Herpaderp came out strong and aggressively attacked SS. Although Will was careful to not repeat his mistake from the Oct 10th event, he accidentally flipped himself by turning too sharply and was stuck in the corner. By this time, SS was limping. Both drive shafts were bent and their wheels were missing chunks. It managed to hobble around enough to beat Dembitz and Ginger Baker but was no match for the battle-ready Mondo who was likely seeking revenge for the KO at Franklin. By the end of it all, SS had some nice dings, 2 bent shafts, broken hubs, 0 wheels, and a belt blade.

But seriously guys, it still works. I said RETIRE it not disable it. :/


DDT (5-0, 1st place)

This time around I had the privilege of testing out DDT's frame against more spinners. Usually because of the design, DDT seldom gets hit on the frame but this time he met a horizontal spinner with a longer reach. Despite the questionable trades, DDT managed to pull out the win in each match and the win overall.

  • Win v Sisu
  • Win v Spare Parts
  • Win v Petit Gateau
  • Win v Revy
  • Win v Spare Parts

DDT's arch enemy this time around was definitely Spare Parts. In the first match I got a lucky swipe and took out a drive wheel. After that I carelessly charged into his weapon head on to "test the durability" of my front armor. This ended up being my undoing as his longer reach eventually broke through the belt guard and snapped the belt. At that point I thought I had lost but to my surprise the judges deemed his immobility more valuable than my lack of weapon. In the finals I decided to be more cautious by avoiding front hits. I knew my side armor could sustain blows and used those to take the blunt and then turn in to slice his unguarded wheels. Soon after DDT had won by KO.


Concluding Thoughts

Even without the v2 upgrades, the CF/Nylon printed v1 frame proved more than sufficient to withstand the big hits from some of the best Beetleweights in the North East. In addition, Silent Spring's performance showed how effective the SimonK firmware ESCs can be in weapon spinup; You'd better believe I'll be trying my hardest to get the Spider ZTW ESC into the v2 design as it may be my new favorite ESC over the Turnigy Plush.

The next MASSdestruction is coming soon on Jan 23. This time it will be hosted by WPI and its unknown as to whether I'll bring new robots or any robots at all. Stay tuned for more. Bot on!

Thursday, October 29, 2015

NERC Franklin Institute 2015

Saturday, October 24th, we left in the dead of midnight to make the ~5 hour drive to Philadelphia for the NERC Franklin Institute robotic combat event. Although the competition supports up to 30 pound robots, I brought only DDT and Silent Spring because robots larger than 3 pounds were prohibited from using LiPoly battery chemistry. Even still, some of the best insect-class robots in the northeast were scheduled to attend and I predicted this would be the real test before Motorama in the mid-winter 2016.

Since MASSdestruction two weeks ago, the robots received minor upgrades. DDT at minimum received a replacement motor can and pulley since the last one was actually bent from years of abuse. DDT was also set to receive a new frame printed with carbon fiber instead of kevlar as the tension member because it was noted that the weapon shaft bore began to open up from all the spinner-spinner hits.

Unfortunately, the prints failed due to some trickiness in the CF thread and I whipped up a quick solution using two belleville washers on either side of the frame to distribute loads more effectively. Silent Spring felt the success of the upgraded drivetrain but suffered some difficulties in the tournament due to the scarcity of the choice motors. Now Silent Spring will run a OEM option, the Pololu 9.7:1 HP gearmotors. The leftover weight was spent to upgrade the disk. Thin sections that bent at MASS-D were beefed up which brought the disk weight up by 4 oz to about 16 oz.

For an added bit of fun, I decided to use some voodoo magic from the robot gods in this new disk. We were clearing out the crate where Overhaul lives and I decided to pull out the AR400 plates we salvaged at BattleBots off Witch Doctor. While one plate was pretty well bent by Nightmare, the other was relatively flat which made it a fine candidate for a blade. I never did get that green color palette from Mike so I ran it in AR plate's natural color: rust.

At 7:45 am we arrived in the museum and grabbed whatever tables were left. The arena was HUGE and pretty awesome in construction. Double doors for entry and exit, a nice kick plate, and great visibility all around. It left me wishing I brought the 12 and 30.

This time around, DDT drew only direct-drive brushless drums. The release of the new Kitbots "Saifu" kit was probably a large factor in that. This turns out to be advantageous to DDT because the drum standoffs pose a big weakness generally to undercutter weapons. As long as DDT can stay clear of direct hits from the drum, the match should go in his favor over several small hits as the standoffs wear away. As a result, DDT went 4-0 for 1st place in the Antweight division

W vs Saifu

W vs Poco Tambor

W vs Revy

W vs Revy (finals)

I was feeling pretty good about Silent Spring's destructive capability after MASSdestruction but utilization of that power ended up being its undoing. The unevenness of the floor and massive gyroscopic forces at high speeds meant that stability was low and its first two opponents took full advantage of that. Luckily, enough, Silent Spring was able to survive those outcomes and climb its way through the bracket. Silent Spring clawed its way through the loser's bracket to find 2nd place in a 5-2 record, losing in the finals to Speed Wedge 3, an extremely robust opponent.

W vs Mondo Bizzaro

L vs Project Darkness

W vs Best Korea

W vs Grande Tambor

W vs Trilobite
W vs Project Darkness
L vs Speed Wedge 3 (finals)

Silent Spring definitely managed to bring the pain just as he had at MASSdestruction. Below are some pictures of opponents after matches.



However, the robots didnt escape unscathed either. DDT lost a rear corner screw when Saifu clipped the edge during a gyro dance but otherwise remained unchanged.

Silent Spring suffered a fair amount of injuries from the first two matches and managed to keep it together for the rest of the tournament. In its first match vs Mondo Bizzaro, SS took a blow to the wheel axle which bent the shaft (and made the disk angle weird) but also seems to have seized up a motor. This caused one of the Vextrollers to fail and I ultimately had to use a postponement to replace the drive electronics. But it was the bent axle that proved to be the biggest issue going into the second match vs Project Darkness. Because it was slightly bent, the disk was no longer planar and so I could not spin up without tilting. Ian is an expert driver (in fact he won best driver award!) and took full advantage of this; he drove around SS while it was uncontrollable and delivered some nice hits deep in the UHMW back and got some nice shots on the drive pulleys which ironically bent the OTHER side's drive shaft. One of these hits actually snapped the wheel hub but somehow continued to drive just fine. I dont remember exactly how the match played out, but in the end PD was the victor and SS had to climb through the losers bracket for a second shot. When I was servicing both bent axles I managed to find a sweet spot where the blade was once again flat. Pretty much "OMG DONT TOUCH IT EVER". In its battered state, it was able to deliver the big hits and crawl back up to the finals without accumulating extra damage.

DDT can continue to compete as is but Silent Spring needs significant repairs to bring it back to a reliable state. Since it keeps doing so well, I must once again delay its retirement even though I want to build something else. I'll assess the extent of upgrades to be received in time for the next competition, MASSdestruction, in December. See you all there!

Monday, October 26, 2015

MASSdestruction at the Artisan's Asylum (Oct 10)

Not too long after the Dragon Con bouts, the crew went right next door to the Artisan's Asylum in Somerville, MA to fight at the first ever MASSdestruction event. As the date came closer, more and more robots registered which helped fund the brand new arena that Rob Masek put together. It was shaping up to be a world-class event.

I brought DDT and Silent Spring with me; neither bot had major upgrades since Dragon Con. Silent Spring received an updated drive train which in turn required re-fabrication of a few critical components. I decided to dump the 33:1 Spark gearmotors in favor of my KW hackmotor which are the drive units I used in Dominant Mode. You may recall, they consist of the KW beetle motors (which are now discontinued) mated to a ~10:1 spur gearbox from the "1000 RPM" gearmotors. Not only is this combination lighter, but it is also faster and more powerful (1500 RPM @ 12V). This change gave Silent Spring the agility it desperately needed to be competitive and deliver the big hit. These tests on an empty Stratasys material cartridge are evidence of its new destructive power.

Saturday morning, Lucy and I hitched a ride with Will and Max Bales who brought a few robots of their own. These included the last minute Hyperderp, lovingly named Turd Ferguson, and Hypershock as seen on TV. All our bots combined, we had 10 ants and 11 beetles. To keep matches rolling consistently throughout the day, Rob proposed a modified swiss tournament structure. All robots will have ~4 matches and points will be awarded for the conditions of the match.

  • A win is 4 points
  • A win by default (no-shows or forfeits before the fight) is 3 points
  • A loss is 2 points
  • A double loss (where both robots stop functioning) is 1 points
  • A no-show is 0 points

The top four robots in each weight class are then pitted in a semi-finals double elimination scenario to fight for 1st, 2nd, and 3rd.

In the end, DDT had 6 matches and went 4-2 for 3rd place. Silent Spring (in a weird turn of events) had 5 matches and went 4-1 for 1st place. The tournament results are hyperlinked here. Only a few matches have video from what I can tell. I will update this page periodically with embedded video as more pop up.

Overall, I am satisfied with the performance of the robots. DDT performed well (never stopped!) but was shown some glaring weaknesses to particular wedge designs. In the future, I'll need to practice driving and prepare special blades to handle these types of wedges. Silent Spring STILL suffers from random shutoffs which is incredibly infuriating but before it fails it delivers some of the most impressive blows I've ever seen in a beetle.

Before Franklin Institute, I'll be looking to redo the wiring and improve its overall reliability. One area that needs improvement is the blade durability; in this event I managed to bend both blades I had for Silent Spring not out of plane (say from a vertical spinner) but IN PLANE. This means its own hits are so hard it damages the disk. I would pay $$$ to see some high-speed video of the blade warping under impact! Stay tuned for 10/24 to see how these two bots shake it up at Franklin!