Archive for the 'Project Build Reports' Category

 

The Boom and Bust Cycle of Chibikart: BurnoutChibi

Mar 18, 2013 in Chibikart, Project Build Reports

All of my projects go through periods of ups and downs. I either end up demoing them too hard, or they get “temporarily” parted out for some valuable component only to never get put back together, or the weather stops being terrible. Given that the winter is letting up and the sun has emerged above the horizon, it’s time to think about the small silly electric rideables once again.

Almost a year after its construction, the story of Chibikart has played out not unlike many Hollywood celebrities and pop icons. After a meteoric rise to fame based on a single hit. which spawned a wannabe or two, Chibikart bungles a live performance but pulls it together just enough for the next show (At the bottom… though I still stand by everything I said in that dribbling polemic). Afterwards, however, Chibikart is never really the same, and sinks into a life of decadence and smoking magnet wire enamel. With my desire to never rewind those damned hub motors again, it seemed like Chibikart was doomed to live under a table forever:


…with his washed up co-stars…

But another season of 2.00silly-rideable-thing dawns, and as such, I must have an Instructor Vehicle. The Instructor Vehicle is totally out of spec and over budget, because Instructor Privilege. I had to decide what to do with Chibikart. It was easy enough to start on another bunny-brained scheme since I have more than enough parts to conceive any rideable, drivable Eldritch abomination, but when I had the semi-working hulk of one of my projects sitting right there taking up space, I found it hard to justify anything else.

So what do I do with it? This was an equally hard decision. I could just rewind the hub motors one last time and be done with it. Chibikart is actually quite fun in hub motor form. I could make, instead, 4 kitmotters to demonstrate the concept in real life even better – one of the ideas on my Infinite Time*Money Back Burner is to make a to-spec Power Racing entry using Kitmotters, since machine time and labor are not budgetable items.

One of my favored ideas from the beginning was to show how to do a sensorless drivetrain “right”. I spent much time in 2.00gokart scaring students away from using R/C style controllers because of their notorious unreliability and lack of current limiting and start-from-zero predictability. Yet at the same time, I wrote an entire section on this in the latest Scooter Power Systems instructable.

The way I see most people use R/C motors with sensorless ESCs in vehicles is in direct contrast to the guidelines mentioned in the Instructable. They’re generally the biggest (or bigger than necessary) motors you can buy, hooked up to a low or moderate gear ratio calculated from load necessity or cruising speed. While this is perfectly fine for DC motor drivetrain design and also sensored BLDC/AC motors, with R/C controllers it’s just asking for trouble. Huge current bursts will be drawn during the starting regime because the motor might not “start” in the right direction, or if it does, needs to expend tons of power to produce the torque needed to accelerate. More often than not, the first thing that goes is the motor controller, since they’re invariably small and unprotected.  It’s been my assertion for a while that a 50mm class R/C motor (typically 1500 – 2000 watt) is more than enough power, if used correctly, to move a person, but usually I see people gunning for the bigger 63mm and 80mm motors. Hell, that’s even what I run on Melonscooter. And precisely like I warned against, I go through motor controllers like crazy and really shouldn’t try using all 6000+ watts of power – it’s literally enough to get me to highway speeds and I have enough fun already flying by parked cars at 20-25mph.

So, perhaps a more noble goal is to troll everyone by making a properly proportioned sensorless R/C power system with typical hobby parts. I’d pair a 50mm motor with an oversized (for overhead) R/C controller, but instead of gearing sensibly and picking the slowest (highest torque per ampere of current) motor, I’d head the opposite direction: An unreasonably fast motor coupled with a double-digit gear ratio. Remember, R/C controllers love to spin things which are going really fast. In the combat robot world, this technique has a nickname: “Gene-ing” it, based on the habits of one particular builder to overvolt small outrunner motors dramatically and gear them down very low to get tremendous power into his weapons. The whole idea would be to divide down my apparently inertia to the motor so far that even the jerk the controller gives it is enough to move me a little.

I next spent some time on two sites. The first was Hobbyking, shopping for potential motors in the 50 to 60mm range. The second was the still-very-useful Torque and Amp-Hour Calculator whipped up for roughing out Battlebot drivetrains over a decade ago. I was aiming for a few goals, in no particular order:

  • Top speed of about 20mph at my desired system voltage of 37v, to try and conform to at least one of my own rules!
  • Gear ratio in the 10 to 20:1 range. Something easy to accomplish with  two relatively simple stages – such as 3:1 and 4:1.
  • Wheelspin current of about 100 amps or less. This was important – I wanted the wheels to actually break traction at a current which I could find an R/C controller to run ‘continuously’. I mentally derate these things by a factor of 2 or more when juggling numbers. This means that the wheel should never just fully stall out

What this came down to is if I apply a step throttle input to 100%, the motors will have a high enough mechanical advantage to start consistently each time, and as soon as they do, they have enough torque to just break traction.

That‘s my current limit. Doing a burnout. That’s how it should be.

Hence, this project was named BurnoutChibi.

motors

At the end of the night, I arrived upon my component choices:

On the right, the NTM 5060/380kv. This was pretty much the only thing available that satisfied my requirements – any slower and I was going to go back into undergeared startup nightmare world, and the few motors that ranked higher in RPM/V in this size were very expensive (by comparison) “competition” motors. The cost of the motor? $36.

I’ll be amazed if it stays together.

On the left is a Castle Creations Hobbyking “160A HV” controller. Gee, it sure looks an awful lot like an older Castle controller – I’ll call it the Sand Castle!

These things have piqued my curiosity for a while, but I’ve never really bought one to chop up. It turns out they are thermal epoxied together anyway, so it’s not like I can chop it up without permanently breaking something. I blame Apple.

While I could have gone with my usual trusty Sentilons (which I know the operating characteristics of well, have decent startup ability, and enough FETs per bridge to let me sleep at night), I decided to take this opportunity to try something new. If the Sand Castles fail, I have 4 left over Sentilon ESCs from the late and great Deathcopter project. The reason I’m running 37v (10S lithium polymer cells) is because I have two giant 5Ah 10S “sticks” left over from the same project. These Lipos are big enough to beat someone senseless with and probably light them on fire at the same time. You R/C aircraft people are nuts.

Two of those packs ought to be plenty to feed two NTMs flying at top speed. Oh, I’m sorry – did I mention I got 2 sets of the above?

tires and brakes

To dump multiple thousands of mechanical watts onto a Chibikart drivetrain would result in the tiny, stubby caster wheels being ground into a pile of elastomeric powder. I’d need something vulcanized and beefy to stand the increased power, but I also didn’t want to go to 8″ pneumatic tires which are the most common size for small EVs. It just wouldn’t fit the character. Chibikarts have to have small wheels.

I decided to try and get some 6″ pneumatic caster wheels. The 6″ pneumatic wheel is something which should exist, but is difficult to find; when it is found, it’s generally expensive and part of a caster already. Harbor Freight doesn’t sell any, and McMaster is generally unhelpful and ambiguous about which one of their 6″ casters is actually pneumatic. For the record, I have a sample of 2717T41, which became the basis for my hunt. I figured if McMaster sold it, there was a cheap and generic verison of it somewhere.

The answer came from Northern Tool, a  Harbor Freight function-alike of somewhat higher market segment. They have a 6″ pneumatic caster wheel replacement for Somewhat Less Price. I went ahead and snagged 4 to see if they were identical.

Also pictured above is a 90mm drum brake I got from Monster Scooter Parts. One of the missions I set out right away to complete was to find a good front wheel brake solution. While both Chibikarts have had rear brakes, this basically precludes the chance of doing any burnout whatsoever. The last time I could do a brake standing burnout was LOLrioKart, and I kind of miss that.

I started out trying to find a small enough disc brake, since I prefer those when I can fit them in. The smallest disc brake setup available stock is 120mm – a bit under 5″, which would have been too close to the ground; I’d be riding on brake rotors if the tire ever went flat or someone bigger than me got on this thing. While I could have just machined a custom brake disc, I decided to be adventurous and see what a cheap scooter drum brake is like. I have plenty of experience with cheap scooter band brakes on front wheels from LOLrioKart’s first front brake attempt, which ended with me discovering what the Capstan Equation meant in real life (Band brakes grip extremely asymmetrically with respect to direction!)

I settled on these and got some samples because I haven’t ever seen them before and was interested in exploring a new part For Make Better Glorious Nation of Silly Go-Karts. It’s my understanding that drum brakes also grip asymmetrically, though to a much lesser degree than band brakes.

As luck would have it, the two spanner wrench holes on the drum line up pretty much perfectly with the hub lug nuts on the wheel. Well then – that pretty much seals the deal.

What I’ll probably do for attachment is make two standoffs that replace the lug nuts with posts I can screw the brake rotor in from the other side with. Two of them will have a pilot “lip” to seat in those holes. I’ll shim up the bore to be concentric, then drill the other two holes using the wheel itself as a template. I’m sure two standoffs might work, but using all four possible locations is just more robust.

This is what the assembly would look like in real life. Instead of using the long built-in torque arm, I’m going to cut it off for compactness and instead mount the brake body itself with a circular bolt pattern. There appears to be enough space between the stamped housing and the brake shoes to drop some #10 or 1/4″ button-head screws into.

I went ahead and made a representative model of this assembly in Inventor. The bore of the brake housing is 14.5mm, so to use with a 1/2″ axle, I’d have to make another spacer with a seating lip.

The next thing to do was to take Chibikart 1′s CAD file and rip everything the hell out of it. Structurewise I was going to turn this into a DPRC, which I engineered a more robust brake pedal for.

I started adding in the new wheels and testing steering geometry, but soon ran into the curious issue of Chibikart being wider than it was long. The pneumatic wheels and brake combination were threatening to push the width out to 30″, because they were just so much wider than the skate wheels!

Some innovative compacting on the steering linkage was going to be needed.

I pushed in the “ears” on the frame a half inch or so, such that the flanged bearings were almost against the 80/20 side rails. Instead of a DPRC-style spindle which is a bolt embedded in a block made of stacked layers, I went for ultimate compactness and designed the spindle blocks to be made from 1″ aluminum square. Like the current arrangement on Chibikart 1, the ‘axle” is a bolt which is tightened into this block.

The hard part was the steering linkage itself. Part of the reason why I made the ‘ears’ so far out in the first place was so I could get enough linkage travel before the ball joint hit the frame. With scooting the steering axis inwards so far, I would only have gotten maybe 15 degrees of wheel travel in either direction. Besides seriously messing with the steering linkage geometry, the other solution was to move the linkage out-of-plane with the frame. I decided to try this first. Basically, the spindle block is solidly connected to the kingpin (which is now a hex head bolt), and on the other side, a big crank arm with a hex bore cut out of it sits like a wrench over this hex head, retained by an axial screw (not shown). This way, the linkage is moved fully under the frame and I can once again have proper steering geometry without  compromising angle.

This configuration might in fact offer too much angle – there’s not a physical hard stop in the system any more, so I might actually add something else later to prevent linkage overextension.

Back on top, I spent a while editing this sheet metal part in-assembly to get the lineup of the cable anchors correct. This plate is essentially an extended version of the “endcap” found on DPRC’s front wheel spindles, with a bolt pattern that engages the spindle block. The mounting bolt pattern for the brake housing, not yet shown, will also be put into this part.

I whipped up a quick 4.16:1 gearbox using VexPro gear models. Vex has gained my interest immensely after debuting their new line of “bigger bot” parts this Spring for the 2013 FIRST season, ostensibly to compete with the likes of AndyMark because FIRST is now big enough to sustain two parts houses. I actually have a VersaPlanetary box bought completely for the purposes of dissection, and intend on writing it up for Beyond Unboxing one of these days – they’re nice, I must say.

The NTM motors have convenient 8mm shafts already, so any gear that can go on a FIRST CIM motor can be fiddled onto this one.

The new rear end of BurnoutChibi nee Chibikart 1 is basically a DPRC back end with a different motor adapter plate. This time, the 4.16:1 gearbox hovers above the main 3:1 chain drive, with the usual chain tensioning method. At this 12.5:1 ratio, the top speed is right at 20mph and the per-wheel burnout current is something like 80 amps. Not bad, but not exactly good either.

So right now, Burnoutchibi is essentially DPRC on some serious roids, with pneumatic tires. By itself, no matter what, it’s not that exciting any more – I’ve built enough things like this already. And by this point, it’s maybe a little hackneyed. The next post will be about what I intend to do about that…

The Soft-Launch of DeWut and the Motorama Robot Conflict 2013 Recap

Feb 21, 2013 in Bots, dewut?, Events, Überclocker ADVANCE

So about that Motorama liveblog…

Anyways, now that the event is done and everything has settled into normalcy again, and with the completion of the best user manual / instruction guide I’ve ever made (I think so anyway…), I’d like to make the DeWut publicly available. Get yours today!

At Motorama, 8 of them were run in various fashions. The three in Überclocker, as well as five in a revamped version of Blitz. Five 3-pound motors in a 30 pound bot. That thing was made of motors. Moto was a great durability test of the gearboxes and outputs under various loading conditions. In Clocker, they were indirect driving wheels and gear-driving the fork. In Blitz, however, they were each direct driving a hard rubber wheel. One gearbox grenaded at the event when Blitz took a pneumatic flipper directly to a corner and bounced a few times. Clocker’s fork drive held up great, to my amazement, because there were points during the tournament when I was basically using the fork as a hammer.

DeWut is one of the two love babies I’ve been working on for the past few months (RageBridge being the other…) and it was great to see my parts actually being able to stand up to some use. Speaking of RB itself, I had no issues at all with my two boards on Clocker, but a few other folks were running the beta version and I recovered some of those when they succumbed to strange issues, to be diagnosed. Blitz also lost one production board to suspected thermal overload (from driving two DeWalt motors in parallel with the current limit maxed out) and another due to possibly a metal chip short from drilling the frame. Another bot’s suffered some kind of strange failure where the board itself looks totally fine, powers on fine, but never exits failsafe mode no matter what radio is connected! I’ll diagnose all of those and hopefully find that there’s nothing seriously wrong with my hardware.

Moving, on, here’s what went down at Motorama 2013.

(more…)

Überclocker: T minus 1, the Opponent Threat Assessment

Feb 15, 2013 in Bots, Überclocker ADVANCE

As Motorama 2013 draws closer, and for once in a long time I actually have a robot done and tested, dammit!, I’m going to do something that I have not done since before this site went up in 2007. I call it an “opponent threat assessment”, and it… is pretty much that. It’s me sizing up the other entrants in the class I am competing in, based on their BuildersDB registration into, and thinking about weaknesses and strategies. I used to do this all the time back in my early battling days, but in recent history (some time around late 2006) I pretty much stopped thinking about it.

Well, that’s about when I stopped winning anything too. Hmm…

The cool thing about the OTAs for me, in retrospect, isn’t the planning and strategizing, which is something that is clearly susceptible to some cursory words and scribble without any real thought put into it. For me, when I did this often, the best part was simulating the match in my head – enacting various scenarios and responding to them. I’d started way back in 2001-2002 with pitting my favorite Battlebots (when Battlebots The Show was a thing) against each other. Watching hours upon hours of videos from local and builder-run events, too, back when broadband was still a big deal, also contributed to building up the models. In this way, by the time I got to high school, I’d already developed a fairly good mental ‘physics engine’ of sorts, since I thought about these match scenarios so often. I usually can, with fairly little effort, stare at a mechanism or mechanical implement and understand how it moves and how it would react to loads. It’s like a mental real-time FEA.

That’s one of the skills which I regularly wonder how you teach – my general opinion is that an innate understanding of mechanisms isn’t possible without many years of practice experiencing them. With each working or broken device you build up the physics engine better and patch holes in your reasoning.  That’s how you eventually get to the point of just staring at something intensely and then knowing if it would perform in a given scenario. It’s also how I CAD – staring intensely at the computer screen while I run through maybe dozens of iterations of a design in my head before putting anything down on the screen (which would take much longer, so by the end of the process I’d probably have forgotten why I built a part).

All that is nice, but besides the point of this post. It’s just something that I wanted to get out of the way since in my recent forays into teaching and TA’ing mechanical engineering classes, I’ve realized that most peoples’ grasps of mechanical engineering concepts are superficial and very reliant on “monkey see, monkey do” kind of copying, or even worse (to me, anyway, perhaps not to my more analytically inclined colleagues) on extremely meticulous and detailed theoretical analysis which ignore real-world implications. While any method could provide a path for advancement and evolution, one of my goals while I’m here is to get undergraduate students to take charge of their own learning and build more things so they also build up strong mental analytical engines.

Anyways, without further ado, here’s the first OTA I’ve written down since maybe 3 or more website iterations ago. The basis is what’s available to me on the BuildersDB for Motorama 2013 in the Sportsman’s class, discussions on the NERC forums with fellow builderrs, videos of the bots in question from events past, or the odd I’ve-fought-this-guy-and-lost experience. This is also assuming the bot doesn’t just fall victim to One Loose Wire syndrome early on…

1. Blitz.

Threat: Moderate

Despite not being pictured on the DB, I know everything about this bot since I’ve pretty much seen it built in front of me. Blitz is in the interesting position of also running RageBridges with DeWut?!s. Why? Because Adam and I are really the people behind the pile of stuff on Equals Zero Designs. The whole damn thing was basically started as an excuse for us to get better parts we couldn’t find elsewhere. Clocker and Blitz are therefore very well matched in speed and tractive force. Blitz’s weapon is a Sewer Snake like dual-hinged flipping arrangement that can throw opponents forward and over (see its first version build midway down). I’d say that Clocker is vulnerable to any attack which can flip it over, not because it’s not invertable, but simply because rolling back on to all 4 wheels takes a precious few seconds. I’d have to avoid being broadsided – a position which Clocker has no defenses against, and I can get continually pushed around in. Because the lifting forks extend out far ahead of the bot and Clocker is known to be very stable even with a 30lb opponent hanging off the fork, a head-on attack might even be my best option. Blitz is fully invertable, but the doubly-hinged weapon would hinder mobility if it’s upside down, a position which I could try and maneuver it into just by using the fork as a flipper. The greatest threat comes from its speed, which is greater than Clocker’s by about 25%, and the widely-placed lifting fingers, which can easily result in a broadside attack if I’m not careful.

2. Diabolical Machine

Threat: Low

DM is a bot I’ve battled before with Clocker in 2010. For this year, the description on the DB reads “Going back to version 4″. Through investigating the builder’s website, “version 4″ is in fact the bot pictured on the DB, and its weapon is a “reciprocating spike”. Besides spikes being actually an ineffective weapon in the combat robot universe, the bot itself is also rather boxy and has no other pushy features like wedges or lifters, and apparently poor inverted performance. I’m anticipating a match filled with much grab-and-go, since its ground clearance also appears rather high. Based on the published build pictures, the drivetrain is not as powerful as Clocker’s, and will probably max out at around 12-15 miles per hour, typical of most cordless drill drivetrains. So short of a spontaneous system failure, I anticipate being able to both outmaneuver and dominate traction. I’m hoping to execute Clocker’s fairly well known spin move with DM if given the chance.

3. Gigarange

Threat: Moderate

Gigarange is a bot I’ve seen in action personally and on video, but haven’t fought. It’s a classic 4 wheel, low profile, 4-bar lifter bot, similar to Test Bot except less wedgy due to the Sportsman’s class rules. Based on the most recent videos of Gigarange at the Franklin Institute event, it’s quick and maneuverable, but I think I have a speed advantage. Its front lifting plate is much narrower than the span of my forks, and the robot is overall boxy and low. I’m fairly certain I can get ahold of it through a frontal attack only. Again, as with all pusher-lifter opponents, I’d want to drive to avoid a broadside attack, but because his lifter is fairly narrow, I may be able to escape from it by rotation – that is, just driving quickly forward and backward if I begin getting pushed sideways. One weakness of Gigarange I’ve observed is that the lifter is fairly slow to act. Hence, again, I may be able to avoid traction breaking using speed alone. The basic strategy would be to attempt to flank to avoid the lifter arm, but if that fails, try attacking full frontal using lift only (to break traction). I’d want to not plant the upper clamp arm on top of its lifter because it can extend with enough force to potentially damage the clamp and actuator (the clamp arm having almost 8:1 leverage on the actuator).

4. Jack Reacher

Threat: High

I’ve been watching the progress of this bot on the NERC forum for months. The bot is one of the few new flywheel powered flipping weapons around, and despite its complexity, the builder is known for reliable designs. Based on test video posted recently, the flipping weapon definitely has enough punch to potentially 360-flip Clocker on a good shot, but more likely, it will just toss me over. I’m rating the bot high in threat just because it can flip and drive reliably (based on my assessment of its drive motor choice and wheel choice/mounting method), which can be bad news for me if I get bowled over and cannot escape in time. Conversely, the complex flywheel machinery may make for vulnerabilities I can exploit by bringing the fork or clamp down on it. The flipper’s geometry is also one I can exploit – instead of “popping out” like many designs, it hinges back such that the majority flipping action occurs at the end of its lifting plate. Hence, if I were risk-oriented, I might actually try grabbing it by the flipper since the actuation motion would try and kick the robot up, rotationally, instead of flinging. My course of action would be to try and bluff the driver into triggering the flipper into an empty shot (e.g. by attacking, but retreating quickly), upon which I would try to either get under the whole bot or attempt to lodge the clamp arm in the flipper. JR can self-right, but with difficulty and only by propelling itself along the ground a few feet based on its test videos, so if I can trap it backwards and upside down in a corner, it will have very little recourse. The worst case failure mode is being flipped upside down, but I hope to be able to recover from the position before JR is able to reload (a process which takes a few seconds as it spools up the flywheel).

5. K-onstant

Threat: High

I’m unfamiliar with both the bot and the builder, and the CAD image posted on the DB is not too helpful. If it’s as described, then I’m going to have to watch out for the spring powered hammer. Clocker’s top armor is definitely a bit deficient, and there are some components sticking up unarmored such as the clamp motor and perhaps the big fork gear itself. Without knowing anything else about how K-onstant drives or loads the hammer, I cannot really make an accurate threat assessment about it. I’m rating it as a high threat because in the event it does work great, I will need to spend most of the match playing defense to avoid the hammer. Against hammer type opponents, I really can only rush them while the hammer is cocking or reloading, hoping ideally for some kind of broadside or up-ending attack with the fork. The best case is trapping them upside-down, with the weapon fired, preferably against a wall, so they have the least chance of being able to self-right

6. Laserbeam Unicorn

Threat: Low

LU is a design I am unfamiliar with, and I don’t know the builder either, but it does have a fairly comprehensive CAD rendering on the DB. The trouble is that I am not very threatened by said drawing. For a lifter, its wheelbase is awfully short and its ground clearance appears limited. The lifter, unlike Gigarange, also does not appear to run the entire longitudinal dimension of the bot, so there is plenty of space for me to plant the clamp onto. The bot seems rather easy to high-center and break traction because of its very short wheelbase compared to bot length. The only issue would be if it were very fast and well driven – but even so, I think I can approach it head on and break its traction with the inner fork tines first (if the bot’s dimensions are roughly what I think they are.

7. Nyx

Threat: High

I think I fought Nyx at least 5 times at Dragon*Con 2012. The match will be completely dependent on driving – the two bots are essentially 1 for 1 in speed. Nyx had a unique ability to wedge itself using its lifting spike between the fork and frame of Clocker and prevent me from lifting, but at the same time trapping himself on the fork. The match will also be dominated by who has better traction as a result. Because of his lifting spike, I can’t approach him head-on like with flat plate bots. Instead, an intricate series of flanking maneuvers (see all Nyx matches in the D*C2012 video) will be needed to get the forks under him. I’m counting on the arena being enclosed this time to hopefully up my unpredictability in maneuvers and intend on using the walls and corners if possible. If I can hook one of his fairly wide and open side rails with the fork, then I have more leverage as a result.  I would have to drive to avoid broadside and rear attacks especially – Nyx has fit very well exactly behind Clocker in the past, and with this build not having changed widths all that much, it will still be a vunerable spot.

8. Palindrome 30

Threat: Low

I watched this bot being built on the NERC forum, and I’m not really sure if the rail of saws will do much damage. Unlike many newbies’ beliefs, saws aren’t that effective in combat because to do damage, you need the opponent to stay still, something which rarely happens. I do expect that Palindrome can do the most “flesh damage” to Clocker, since spinning saws are spinning saws, but unlike many other opponents it has no capability of pushing or wedging. The weapon is also driven by an easily stallable brushless motor and runs in solid bearings, so it could bind very easily. The strategy with Palindrome would just be to grab and go. I do hope to parade him around the arena and mark up the walls or floor.  The bot also has a broadside vulnerability something which I hope to be able to exploit because its speed potential does not appear to be great (using DeWalts in low gear, though with large wheels).

9. Phoenix

Threat: High

Phoenix is a quick and maneuverable flipper bot which I have seen dialed in recently – it was seemingly unreliable in the past, but now it consistently flips 30lb opponents and can also self right handily. Because of the length of its flipping arm ahead of the bot, I’m going to have to avoid any engagement directly, or allow broadsiding. The arm reload cycle does take some time, during which it is raised up, so I could potentially bluff a flip, then attempt to lock the fork under the arm to block him from reloading. Clocker drives much faster than Phoenix, so I should be able to maneuver as needed. Another potential strategy is to keep the fork slightly up, over the height of the body of Phoenix, and attempt to hook his lifting arm at the top where the cylinder attaches. The body of the bot is also short enough to allow a broadside grab. Overall, I’m still rating Phoenix as a high threat because of the potential to flip Clocker over handily if I miss a beat.

10. Such and Such

Threat: Low

Based on the previous version of S&S and recent Facebook photos posted by the builder, S&S is again a “horizontal clamper” – the whole bot expands sideways using a multibar linkage in the center, and can clamp down on you from the side. It then uses dominant traction to corral you around. This year, S&S is actually a “shufflebot”, or a pseudo-walker that uses continuous cam legs, with what appear to be rubber blocks for legs. Walkers are afforded a 50% weight advantage, so S&S may weigh up to 45 pounds. While Clocker could lift it, I’d have to make sure to grab him on a long side (so the bot’s weight is not substantially leveraging more than an average 30lber) but that, of course, risks being grabbed in return. I do think I still have the traction advantage, however, and definitely a speed advantage because of his shuffling nature. As a result, so long as Clocker doesn’t mysteriously fail, I don’t think I can do poorly against S&S so long as I keep driving and avoiding the hug of death.

11. That Robot

Threat: ?!

There’s not enough information on the DB regarding this bot for me to really make a call. Allegedly it has “spinning arms”, which puts it already into the gray zone of Sportsman rules. Clocker is built fairly solidly, so if “spinning arms” does become a real thing, I hope to be able to back into them and stop them. Otherwise, the bot’s design sketch tells me it’s not invertable. I’ll have to wait and see for this one

12. Tyrant

Threat: Low

This year, Tyrant returns with an actual chainsaw attachment. Trouble is, I don’t think it will do that much damage – it’s not geared very highly, and like all saws, will probably bump and skip off a moving opponent. However, in the name of the class, it will put on a GREAT show I’m sure! Another one of those perennial n00b weapon suggestions is a chainsaw, so many people in the audience ought to identify with Tyrant. Because it has no pushing implements and big exposed wheels, I’m going right at him. In fact, I kind of want to try grabbing him by the saw. Tyrant is quick, however, and the chainsaw is sure to win aggression points from the judges and audience, so I’m going to have to control completely (Complete Control style!) or risk losing by decision.

13. Upheaval

Threat: High

Upheaval is the bot which has pretty much won every Sportsman contest there’s ever been. It’s reliable, packs a massive punch, and well-driven. It also has front drive wheels, so it can really just turret around and wait for and of your maneuvers. I fought it in 2010 with Clocker Remix to predictable results. This time around, I should have actually functional drive motors, short of a spontaneous failure (which is always a potential factor). Clocker is now much faster than Upheaval, but his turreting means I’ll have to be clever in my approach. The basic strategy would be much the same as fighting Phoenix or any other flippy bot – try to bluff a flip, then get under him while the arm is reloading. Clocker has many apparently solid spots up front which I could use to my advantage – the fork will tend to slip its clutch if a sudden force is applied, and the springy legs will hopefully live up to their name . Alternatively, as long as I can keep rolling him over (not grabbing), he’d have to waste shots having to self-right, and I could potentially try and trap him against a corner that way, making self-righting impossible. As long as I can keep moving and poking, I should be able to avoid being flipped. The most important part for me would be to never, ever drive across the flipping foot and never engage directly.

Basic strategy for Clocker also goes something like:

  • Try using the fork as a lifter first, to leverage the opponent off the ground, then grab only if needed
  • Don’t body slam people backwards – Clocker may not be able to exit this position, requiring a two-robot unstick pause in the match. Maybe only do it for effect at the very end of a match if needed.
  • Drive slowly and methodically unless I need to run – recently my “stick twitchiness” has gone up due to me being seriously out of practice. I hope my practice driving with Clocker has been able to resolve it
  • Avoid being broadsided at all costs – Clocker Remix had that weakness, and Clocker Advance has the same long flat sides.
  • Drive upside-down to escape a flip if needed, don’t try to self-right on the spot.

Hope this all works out! I’ll be leaving for the tournament in an hour or two, and hopefully tomorrow there will be a live report from the event.

Überclocker: T minus 4

Feb 13, 2013 in Bots, Project Build Reports, Überclocker ADVANCE

Continuing on yesterday’s update, here’s the tale of wiring Clocker up over the past two days, plus (finally!) some video of driving and testing. As of right now, I’m pretty much ready to call the bot “done”. As in, maybe not everything is perfect yet, and perhaps not everything has been tested to breaking… but if Motorama were in fact tomorrow, I’d be comfortable with pitching it into the arena.

With the bot basically having reached mechanical completion with the routing of the drive chains, I turned my attention to the wiring end of things. Clocker uses two RageBridges (And So Can You!), one for the two drive channels and the other for the clamp and fork.

In the past, I’ve bussed together the two power inputs of RB on the bottom, but since now I’m working with the boards with heat sinks, this is no longer possible. Solution? Just bus across the top. Now, “production” RBs are supplied with 4 battery wires, but this is one way to cheat a bit and get a one-input system. Yes, I know this layout is terrible.

The green board is actually one of the revision 5 prototypes which are functionally identical to the production boards – I didn’t want to consume a production board which could be sold for my own giggles if I had workable prototype ones.

Next up was making the power source of the whole bot. Originally, I was planning on an 8S2P configuration using A123 cells. However, actually assembling 16 cells together and trying to stuff them into the rear cavity with a RageBridge made me realize that it was just utterly impractical. The fit was extremely tight in CAD, and the CAD model does not include the many layers of shock absorbing rubber foam and heat shrink I coat these things with.

So I dropped to 7S. No matter – that just means Clocker goes about 17mph instead of 20! Big deal, in a 24 foot Motorama arena or a ~12-18 foot Dragon*Con stage!

This is a picture of the pack in progress. I used my usual construction technique of copper braid and split balance harness (so my fairly average hobby-grade charger can actually recognize it).

Like Null Hypothesis, I’m giving Clocker an integrated charge and switch port. There’s a Deans connection which is really just a removable link, then an XT-60 type connector which has a direct battery connection. Why do I make them different? Because otherwise you risk dropping that removable, super low resistance link right across the battery. Yum…

To test the drive, I mounted 1 RB and the battery in the rear of the bot. The “production” RB was used because it has heavier traces than the prototype and the drive motors will be stressed much more, current-wise. With only the drive hooked up, I took the bot for a spin around the hallway.

I’m very satisfied with the drivetrain. Unlike Clocker Remix and Clocker version 1 in 2008, there was no attempt to keep the center of gravity as far back as possible. Rather, it’s near the centroid of the wheelbase and track rectangle. Result? This thing handles so smoothly – almost as good as Null Hypothesis, which still wins just because it has fatter wheels – Clocker tends to drift and slide. Previously, Clocker had a particularly nasty oversteer issue because the weight was purposefully far back, causing uneven wheel dominance in turning.

A better shot of the controller installed in the bot. It’s actually raised off the bottom plate by 1/4″ spacers. The threaded standoffs act as nuts to secure the board against those spacers, and then the fork/clamp controller sits on those threaded holes. The receiver is stuffed right next to everything by the back.

And the ‘upper deck’ in the Tower of Rage is assembled.

The top plate needed a bit of filing, sanding the edges, and enlargening the waterjet-cut pilot holes to slip into place. It’s retained by the same #4-40 button head screws that I normally hate so much, but hate #6-32 even worse, think #8s are worthless, and consider #10s too big for the job. So… yeah.

With all of the electronics and pretty much all the screws in, I took the bot for its first moment of truth: the weigh-in.

Uhhhh… well that isn’t good.

Now, what on earth could I have missed? The CAD model included almost all screws and the full 8S2P a123 pack and still came in at 28.5 pounds! I’m going to just assume that things weigh more than what I could have guessed – for instance, the threaded rods binding the fork together weren’t modeled, which seems silly since they’re quite huge. Additionally, the top and bottom plates were modeled as phenolic material when they are actually fiberglass (garolite G-10). The difference in that alone turned out to be almost 0.2 pounds per plate!

Well then. Clocker has to ditch 1.25 pounds somehow.

I spent a while thinking about possible plans of escape. I had very little metal that I was comfortable with “speed holing” since it would compromise the somewhat complete armored perimeter of the bot.

Replacing said top plates with polycarbonate would save about 0.35 pounds per plate, but I did not have the material on hand at the moment, and it wasn’t enough by itself.

Dropping to 6S on the battery would only save a few ounces, and I’d have to tear the battery apart again. The final solution had to involve multiple changes.

With the heaviest of hearts, I grabbed one of the left over 7S 4.4Ah Thunder Power lithium polymer packs from the amphibious DERPA project team and replaced the A123 pack. Sadly, no matter how much I love the little white round cells, the lithium pack just has greater gravimetric energy density. I gain back what is essentially an 8S a123 voltage (25.9v nominal), with essentially the same capacity. And, it saves 0.75 pounds; even with the two G-10 plates I added to the top and bottom to make the pack simulate a hardcase battery so it isn’t as squishy.

Now, with Clocker at 30.5 pounds, knocking the rest out of metal was a possible course of action.

I always figured I was going to have to machine out these landing legs eventually. They’re solid 3/4″ aluminum. That’s almost obscene. Each slot basically netted me 0.14 pounds, so a cool quarter pound for the pair.

There’s a quarter left, still.

The rest of the quarter came out of the right side outer frame rail, which was solid 1/4″ aluminum. I hollowed it out to a wall thickness of 0.1″ on the outside, and this was able to knock out just under 1/4 pound. Since the Sportsman’s Class does not have to contend with heavy hitting cheap shot kinetic weapons, I was completely fine with this relatively thin (for a combat bot, anyway) side armor.

The final result?

Yes.

That’s as close as I want to go. Generally, big event scales only read up to the 0.1 pound anyway, and some leeway is given at the organizers’ discretion to account for not everyone owning the same calibrated scales. So I should be able to add 5 or 6 more #4 button heads on the top plate without issue.

So, the final tally of changes was a few machined-hollow metal features and a battery change. While the battery change lost me 0.75 pounds in the rear, the front leg channels and side plate pocket combine to keep the C.G. basically where it is.

Finally, the press shot:

I spent way too much time driving this thing around and practicing with the fork. So far, the driveline has been perfectly reliable. The centroid CG placement and smooth braking of the RageBridges, coupled with the ultra-tight deadband, makes for one of the most smoothest-driving bots I’ve built.

The “clock face” has been covered up with some thin polystrene sheeting that is normally used as thermoforming exercise stock. It’s just a dust cover, more or less.

One thing I experimented with was adjusting the torque clutch on the DeWalt gearboxes. My “DeWut?!” mounts have a set screw that can push in the torque clutch’s preloading spring, just like what the torque setting ring does on the drill body itself, so in principle you can create a torque limited drive. I’m glad to say that this is in fact possible in real life. I tried lifting stuff with the fork, but the motor just made angry drill sounds because I didn’t adjust out the clutch at all.

Luckily, this version of Clocker was built with serviceability in mind, and in under 1 minute I had the 4 screws undone, the motor slid out, the set screw tightened, then the motor remounted and the screws reinstalled. While I could have played the “How far do I turn the screw to lift a 30 pound opponent?” game, I elected to just let the current limit on RBs take care of maximum lift load, so I locked the clutch using the set screw.

My exercise regimen was mostly blasting in straight lines up and down the hallway. This was in fact not very easy – Dewalt must have changed their 18v motors’ windings very slightly at some point, because I have 2 allegedly identical 18v type motors that are  definitely different ages, and they are very slightly different. They are different in speed enough, though, that Clocker still pulls a wide circle. For now, I’ve been just practicing it away with the radio, but I probably want to check for matching motors next time I have the bot open. I also engaged in a few short “drive a perfect square” laps, and “do a perfectly straight J-turn” also.

But Clocker has actual weaponry, so I’ve also been practicing attacks and methods using the fork and clamp. I’m fairly sure that now with the much faster clamp arm, I can catch and lift an opponent (or at least break its traction) in under 2 seconds.

Here’s another short test video using Null Hypothesis as a fork chew toy.

Notice how Clocker can hoist a 30 pound opponent pretty much hanging off the end of the forks. I attribute this to the much longer springy legs this time around, making the 2-bot complex much  more stable as a result. I think in battle the lifts won’t be as smooth, and there will be much squirming on the opponent’s end, but it’s good to see that Clocker will no longer be as prone to faceplanting on a lift. The down side, in my mind, is that the legs are now a vulnerability because they stick out so far.

The tournament will soon tell all. Between now and Motorama, I want to get a few things done:

  1. Practice sparring with some of the other area 30lbers, Null Hypothesis, or the now mostly beheaded Clocker Remix.
  2. Make spare wheels. I have 6 spare wheels, and I’m probably going to need all of them.
  3. Secure all internal wiring with some kind of adhesive or sealing compound
  4. Take nice pictures and make an assembly guide for the DeWuts! I plan to bring several to Motorama, so they must be ready by then. They will be up for public sale on e0designs.com after the event itself.

For the T-minus updates from here on out, I’m probably going to just update on practices and anything that breaks, plus maybe an opponents analysis when the event gets closer.

 

Überclocker: T minus 5

Feb 12, 2013 in Bots, Project Build Reports, Überclocker ADVANCE

Since the last Überclocker update, a ton of work has been done on the bot. This past weekend, I purposefully trapped myself in the shop for the duration of the Snowmageddon Snowpocalypse Snowlingrad Snolocaust Great Leap Snowward Snowtorious B.I.G. which shut down most of New England for Friday through Sunday, to get as much done on Clocker as possible. I’m proud to say that at this point, the bot is driving (but not yet lifting). This following report will basically summarize the work of the past week or so, including all of this past weekend – otherwise, there’s going to be like 90 pictures!

I began by tackling basically the only menial machining task on the bot: making the Springy Legs. The “raw forms” were waterjet-cut from 3/4″ aluminum with the intention of finish machining.

This is the stationary rear portion of my DIY shock absorbers. They just have a hole drilled through to hold a bronze bushing, which will interface with the “rod”, a shoulder screw.

The other two parts were a little more interesting. The trunion end of the shock absorber is doubly supported in the leg itself, so it meant I had to cut a roughly 0.8″ deep slot in the leg. I broke out, fortunately not actually breaking, my long-cut 1/4″ carbide endmill from the days when I collected tooling and carved it in 3 passes.

After most of the little menial machining objects were done, this is the state of the frame. I have a set of stiffer springs in case I find these too ‘soft’. If the bot lurches forward too much, it could hinder the lift by just keeling over at the very front of the legs.

From there, I moved to making the wheel hubs. These are based heavily on my design used in Clocker’s summer Gritty Reboot, but with a bigger Delrin center and larger diameter spacers for the larger sprockets. I started with a 1″ Delrin round and quickly whipped them out on Tinylathe from there.

Starting to reach Criticality (where a project can finally support its own weight)… I found a pile of nylon and steel washers to space the legs out properly. From there, plenty of blue Loctite was poured into the standoff-axles, forming the “permanent” side of the attachment. If I ever have to replace the legs, though, I’m kind of boned.

I christened this thing “Clockerboard” since I was riding it like a skateboard for a little while. The frame is extremely rigid, and the precision-ground aluminum standoff axles pair well with the Delrin hubs. The action is so smooth it might as well be on real bearings.

And the bottom plate goes on. I clearance-drilled the pilot holes in the bottom plate and made the attachment with #4 button headed screws. I normally despise buttonheads because of their super small 1/16″ hex key size, but they offer more bearing area (larger head) than the regular cap screws, which is better for the garolite’s structural integrity.

Starting from a 1″ diameter chunk of ceramic-coated (hard anodized, anyway) aluminum shaft, I drilled out the entire center with a long .75″ drill bit to save weight, then started putting the fork asesmbly together. In lieu of using retaining rings on the ends of the shaft, as originally planned, I decided to just machine spacers to span the gaps between forks in order to take up the axial slack. There was already going to be 3 different shaft collars constraining movement axially – a little snap ring wasn’t going to add that much more to the equation, and I could not be buggered to try and find the only grooving tool within 2 miles.

After cutting out some lengths of threaded rod, the fork comes together. This assembly is extremely rigid because of the sheer amount of preload I’ve put into this system through the three alloy steel threaded rods.

…unfortunately, I forgo that screws have heads.

The real story here is that the shoulder screw holding the leg on sticks out further than in the design because I added the spacing washers to give the leg a wider bearing area. I didn’t account for the extra width of a locknut compared to a regular nut, either. So the result, unfortunately, is just the hardware running into itself.

No problem – cutting the threaded rod exactly to length, and shortening the outer spacers 1/8″ each, gave enough slop room to clear the fork.

The modified clamp actuator goes on with, you guessed it, some more shoulder screws.

Mid last week’s pretend-o-bot, or thereabouts. This was just before the DeWuts showed up, so at this point I was stuck until I had motors. Luckily, Anonymous (…Chinese CNC shop) delivered.

While waiting for the DeWuts, I decided to take care of the rest of the menial machining tasks. I purchased a 1/8″ keyway broach to make the key cuts, and went to the Edgerton Shop to use a long-throw arbor press to cram the thing through the gears and sprockets.

With the DeWuts having finally shown up, here’s one of the first pictures from the mid-snowstorm robot work: one drive motor with shaft trimmed and test mounted. To be honest, I was mostly distracted by finally being able to run Landbearshark in its native habitat to get substantial robot work done…

I also went ahead and mounted up the lifter motor. It’s attached to the frame through a big U-bracket which functions as a spacer, the bot cavity being 1/2″ wider than the motor.

One issue was that the final distance between mounting holes was basically 3.00 instead of 3.05″, the original anticipated design length. I’ll have to take measurements of multiple units in order to confirm the +/- deviation from 3.00″ I should report, but the 3 on Clocker are all pretty close. Either way, much clearance-drilling was required. I basically had to open these holes up to 5/16″ to make that fit.

Drive motors mounted. By this point, I think the snow depth was already 18″ and going. The pickup truck plows had given up, and heavy equipment was starting to roll through the streets.

The last task of the night was routing the chains.

The chains on this version of Clocker are routed a little interestingly. On this (right) side of the bot, the chain wraps around the bottom of the sprocket. On the other side, though, the chain wraps over the top of the sprocket (and hence under the tensioner sprockets).

I did the chains up this way on purpose because the DeWalt motors are very heavily timed to favor one direction: the drilling/screwing-in direction, or counterclockwise. You rarely use a drill in reverse, so manufacturers squeeze a bit more forward power out of the motor by optimizing the brush timing for one direction.

In a traditional two-sided drive robot, the motors have to spin opposite directions to effect forward or backward motion since they are mounted mirrored from each other. In the DeWalt’s case, it seems to cause up to a 10% speed difference between sides – that means the robot will just pulll a huge wide turn the entire time you command ‘straight’. Clocker Remix, in fact, does this – it has never ‘driven straight’ in testing.

In practice, combat driving never really sees enough straight line travel for this to matter much, so most people just straight up ignore it.  If I had the opportunity to make the bot more symmetric, though, I was going to take it. So, to power the bot forward, both motors in fact rotate counterclockwise as viewed from their own shafts. This is the favorable timing direction and the difference in speed is both audible (faster spinning, higher pitched) and visible (the bot is definitely slower going backwards).

As for the chains themselves, I incrementally dialed in the tensioners by running the chain for a few minutes, then moving the tensioners to tighter positions. Chains stretch a few % just by virtue of wearing in the first time, so this was critical.

That concludes the first round of work. In the past day, I’ve managed to wire up half of the bot, but that will be reported on shortly. Tomorrow I anticipate being able to start doing shakedown tests and figuring out what to tune before the event.