Category: Overhaul 2

The Overhaul 2 Design & Build Series, Part 2: How to Design an Overhaul!

chorl1 Comment

In the previous post, I presented a summary of why overhaul1 sucked the various shortcomings of the Overhaul 1 design. This will be the first of two (or more…) posts which heavily dive into the details of CAD modeling, and SOON! you’ll get to see the Overhaul 2 design in its entirety. The idea of these posts is to show how the design evolves incrementally from the crude sketch to complete model…. and also the little changes made to it even after it’s “complete”, because that happens every time. Always.

My software of choice is Autodesk Inventor. Yup, I’m a child of the Autodex. They got to me first in 2005 when I was a wee lad high school sophomore, with handing out Inventor Professional to FIRST Robotics teams. MIT largely ran off Solidworks, and during my time as an instructor I also taught intermediate to advanced Solidwanking (that’s a technical term) for the design classes, but whenever I do something for my own projects, Inventor is still my go-to.  These days, Autodesk tends to push Fusion 360 to the maker crowd and it’s generally free with limitations. Even just a few years ago, if someone asked me how to start using CAD programs, it was hard to give an answer that was accessible. That’s changed now with Fusion, 123D, Onshape, and other platforms, which is great. If you’re a student or have .edu affiliations, you can get students’ and instructors’ licenses for Free (as in Beer) from Autodesk Educational Community for a few years at a time.

So anyways, all the screenshots you’ll see will be out of Inventor Professional 2016. My opinion is that on the average non-commercial (or even light commercial) usage level, the difference between Inventor and Solidworks is largely like Ford and Chevy – largely the same features, same commands, and same structure are seen in both, with some nitpicky differences such as I’ve never successfully turned on a headlight in a Ford car, I’m pretty sure. “Solidworks vs. Inventor” is like the vim & emacs debate of CAD students, so I’ll just say that I’ll have none of it here.

Let’s begin with the sketch from July 2015.

 

This sketch was actually the last in line of several I was playing with, which sadly were not saved. It originated from the desire to push the “dustpan” front end of the robot into the drivetrain a little, such that I could have front wheels almost directly under the center of lift. This was something that we played with for Overhaul 1, as shown in this CAD party photo from last year.

As can be seen, the bot was a lot more sloped, and the earliest designs of the shuffler modules reflected that. We were going to make the front armor heavily sloped and stuff the front of the drivetrain under it. But that evolved away for reasons which I do not remember well – possibly ease of fabrication focused.

The rest of the shape of the sketch followed the thread of putting wheels under the new dustpan region. Those wheels had to be smaller than the others, which was fine. I originally had the design as 4 wheels instead of the 6 shown (3 per side), and I think that would have been passable. However, I decided to explore 6 wheels to give me more options for wheel placement – basically, tuning the “front” and “rear” wheelbases independently as needed. 6 wheel drivetrains also have the benefit of being able to place the center wheel very slightly lower (“drop center”, for you FIRST kiddos) such that the bot is basically 4wd at any point in time on a shorter wheelbase, preserving turning speed while affording a large stable base.

Beyond that, the sketch was simply “what looks kind of okay” on the dustpan profile and body length I drew. There was no science behind the other dimensions – I just wanted something on screen to stare at. I liked the idea of a multi-faceted dustpan for cool looks, but wasn’t entirely set on it.

Fast forward to October. Season 2 was supposed to have been announced 3 months ago and I was in the middle of RageBridge 2 development, when suddenly, applications for #season2 opened up! Uh oh, now I actually have to put things in the frame sketch. It became obvious very fast that this bot was too small – nothing useful could fit inside the space as-drawn, even though it looked nice and compact!

So I made it longer.

As you can see, I literally made every dimension a bit longer, and put some sketch boxes in to represent motors and stuff. Also shown in this sketch is an imported side profile of Overhaul 1. Yes, even after stretching the new design (“Holy crap, 3 feet?!”) it was still shorter than Overhaul 1.

This sketch was reasonable to start making “3D” so I could start playing with placing part models.

So I did just that! I took this sketch, stuffed it in an Assembly file, made it a reference drawing, and started extruding parts off it. Just solid blocks.

INTROUUUUUUUCIIIIIIIIIIIIIING

 

OVERBITE!!!!!!!

Sigh.

Everyone who saw this sketch model said it looked like Bite Force. Well gee, I’m sorry not sorry if some times in engineering there is an optimal geometry to get something done. That’s going to start biting BattleBots in future seasons, I think, when more of the design space is “claimed” by certain successful designs. But we’ll burn that bridge when we come to it, as I always say.

Noticeable additions to the sketch include little side strakes on the lifting arms. I put these there for looks largely, but Überclocker’s “fish hooks” have been wildly successful in snaring bots that the fork gets under, so I was planning on including similar entrapment devices. Did I mention pointy bots look cool? They do, right!?

Another side profile shot with OH1 in the background as a halo.

Whoa… quite a jump from the last picture here. Actually not so much. I just hid the square arm and replaced with with 1/8th of a circle that intersected the lift axis and was tangent to the ground, and then extruded a bunch of shapes off it. I think this looks cooler – the curved arm works well with the upper arm (which has been separated into individual pieces, but still all driven off the big assembly master sketch). Not only that, but the constant curvature is stronger than a sharp angle joint with faced with the bending forces imparted into it by the top arm.

I was playing more with the top and back sides of the bot at this point, trying to emulate the sloped-back look of OH1. I made a simple Sheet Metal part with 2 bends in it to give some more non-square shape to the sides. And the cherry on top was of course the now-classic Overhaul Ears, also a Sheet Metal part that was referenced from the top center hole in the arm.

Nothing in this assembly can move – it’s all just sketches and extrudes. It’s a visual clay-model for me to think about the shape and size of the bot.

This is the previous sketch model with the top plate removed. The arm towers just grow out of the big solid rectangular extrude feature that is the body, so my next step was to hollow it out a little bit to add part models inside.

Now it’s starting to look a bit like a robot. I stuffed some sprockets in to get an idea of plausible drivetrain scenarios. How much chain/gear/belt ratio you can use is largely dictated by things like wheel size, and that’s influence by robot shape. I grew this design from “robot shape” downwards – many people grow designs from “I have this motor” upwards. In this case, since it’s a bot which is trying to take after a previous design, I wanted to establish the form first.

Notice also that the bent-style top plate is gone, in favor of a single flat plate that forms the sloped back of the bot. Again, just playing with shapes – I decided to let component placement dictate where I put top armor and back armor. So long as things fit inside, we’re chill.

Look! It’s Sadbot! I have a whole build report on this thing that some builders in the community will find very interesting.

Especially after this next image.

Here I am trying out a few test components. Keep in mind that these were all part models I had on hand, or extracted from manufacturer websites. The blue bricks are batteries – the same 8.0Ah, 18.5V lithium packs we used in OH1, just a flatter configuration. The purple things are motor controllers, and the gray cylinders are motors themselves. These are quite different from what I’m used using in bots, especially big ones.

I can hear jaws dropping, the “I KNEW IT”s, and the finger snapping from some of you already. Again, explanations in due time, I promise.

Also shown in the photo – on the right upper corner, two Whyachi MS2 switches, since the new rules required separating drivetrain and weapon power, and some more big sprockets and gear models downloaded from McMaster-Carr to, once again, get a rough idea of how much gearing I can put into this frame. Less than 2 hours of clicking around separate the last two images, so no magic is being exercised here.

Here is a different configuration of the interior. Motors and controllers are often the biggest-ticket items in a bot design, so you kind of wrap the thing around them because they ultimately dictate how your robot drives and behaves. They’re also the juiciest and most prone to damage from abuse, so it’s good that everything is kept in the center.

Instead of a “2 x 3 stack” like previously, this arrangement is just all the motor controllers in a row, laid out flat. I liked this arrangement more, because it lets all controllers be accessed equally for replacement or service if needed. I imagined just having a huge bus rail attached to the battery pack that the controllers plug into, so I can rip one out easily if need be. Their heat sink fins being vertical also helps for heat dissipation, which if you’ve read enough RageBridge development posts, is the number 1 enemy of motor controllers.

The 7-in-a-line arrangement was very tight on space though. I could have made the bot wider – and did, briefly. But on one side of the bot, I played around with pushing the drive motors further out. The gearboxes shown in the design are Banebots P80 gearboxes, which proved themselves well on OH1’s lifter and clamper/crusher gearboxes, so I contemplated using them for drive. Part of Sadbot’s mission was to see if a few of them could handle a 250lb-class drivetrain. In this sketch model, I had the bright idea that I’d basically use the P80’s longitudinal structural bolts to mount them to the frame – just use longer ones, so they poke out the back of the gearbox.

Pushing the drive motors further out enabled me to free up almost 5″ of interior space across the width. That was plenty to put the controllers in, separated comfortably, with room for future mounting hardware, since you don’t just throw things inside a bot frame and hot glue them in…. Okay, fine, some of you do.

It was in this state that I submitted the application to BattleBots for Season 2 in late November. Everything KIND OF makes sense, everything OUGHT TO fit, and SHOULD work on a good day. Seemed legit to them!

As for how this region ended up – here’s how it looks in the finished design.

well that escalated quickly

Some things changed. Some things are gone. Some are still there! The next few posts will tell the story of how everything got to that point. You know how I said “oh, not much work separated these two images”. Well, a whole lot of damn work separated these last two!

It was now late November. Time to stop CADing for giggles, and start CADing for real. This is where it begins….

s

The Overhaul 2 Design & Build Series, Part 1: How to Overhaul a Design

chorl2 Comments

Here we go! The very first (real) post in the Overhaul 2 Design & Build saga, which will set the stage for the future posts. What I’d like to do with this post is give a little more story on how Overhaul 1 was designed and built, its shortcomings in the competition, and then move onto how I think the new build should address those issues. Then, from here on, the posts will be exclusively design and build oriented. My anticipation is that Parts 2 and 3 will be very long, showing the evolution of the CAD model, and from there, the posts will be more scattered with build updates and interspersed with CAD updates.

The first step in evolving a new design is always to appraise the functionality of the old, and to do that, it’s time to review what made Overhaul 1 into what it was for Season 1. In the first portion of this post, I basically summarize the build process for Overhaul 1, so that’s worth a read. What you never got to see in that post, since it was made prior to the airing of Season 2 episodes, was the competed design and the near-completion bot. We weren’t supposed to reveal anything yet, much like how you’re not “supposed to” be doing now.

So here’s some never-before-seen pictures of the build of Overhaul 1. First, the finished CAD model:

There’s the beast in its entirety, designed using Solidworks 2014. Trying to twiddle this model a year after the fact has caused me to forget much of its hacked-together behavior, and additionally made me wonder what the hell “midline axis of fancy thing” was (it was the central axis of the leadscrew nut on the top clamp arm…)

Documentation. We has one.

Recall the photograph in the “Life of Charles” post which shows megatRON mated to Uberclocker.

You can basically break down OH1 into three components. The front (leftmost part of the CAD model), which is the dustpan/corral of Ron, the upper arm and clamp which is heavily Überclocker, and the shuffler modules inspired from Fission Product. It was largely designed in this manner, with the four of the principals (J, A, me, and D) moving from ‘station to station’ with the design.

We basically knew from the start that nothing about this design would be ‘optimal’ due to the genetic confluence and conflicting design requirements. So the design was kept very modular on purpose. The drivetrain was one box, the dustpan was another, and the arm stuck on top. The positioning of everything ended up being dependent largely on where internal components had to go, which is a theme which will return for version 2. This ended up giving Overhaul 1 it’s very characteristic “hunchback” look in the arena.

Not only that, but the wheeled drivetrain seen in the tournament matches in Season 1 had to be made to fit entirely in the space previously occupied by the shuffle modules, resulting in the wheelbase being very short – only 10″ on a bot which was 41″ from end to end. It might as well have been 2WD!

 

While this made for quick turns, it meant the bot had no front traction centered under the massive front dustpan, and furthermore, was already very front heavy with the lifting arm and its 2 actuators (plus the top clamp arm) hanging off it. The little black and white ball below shows the center of gravity – as you can see, it started past the front wheels.

When we got anyone on the forks, it was often at the expense of being able to, you know move. And then there’s this embarrassing shot:

This, coming from someone whose bot was known to do the “robocopter” was borderline shameful. This goof was a result of putting the large rubber shock isolators (“wubbies”) in a straight line at the dustpan’s interface. We just kept it to the height of the bot frame for convenience.

Contributing additionally to traction issues was the large overhangs of the drivetrain caused by the switch to wheels. The shuffle legs could reach well past the “wheelbase”, but wheels, obviously, were limited to the contact directly under them. This meant that there were plenty of moments where OH1 was lifted up backwards – either by opponent attack or counter-torquing the lifting arm, such as the first Lockjaw match below, where I couldn’t get out of it without lowering (or being lowered) back down.

OH1 could really only be tilted back about 15 degrees before losing rear traction, and this was a key issue in why we lost the Bite Force match (besides my too-aggressive tactics, which is not really a design problem so much as a firmware issue with me). Any time Bite Force got under me with some energy, it was enough to prop OH1 off the ground.

Besides driving, the arm itself was too slow – watch OH1’s matches again and you’ll see quite a few missed lifts. I’m used to the very fast lift of Überclocker, and the linear actuator driven arm was a great deal slower than that, and lack of practice really hurt my timing.

The reasons why we went linear actuator over rotary (e.g. Bite Force and some other lifters like Stinger) was partly due to available parts and partly due to lack of experience. It was our belief at the time that you couldn’t get a gearbox heavy duty enough to handle the torques – many hundreds of ft-lb – without it being massively heavy or requiring another stage or two of torque amplification to use, which also increases weight. A linear actuator is a cheap way to get immense force at the expense of speed. In fact, Overhaul’s lower lifting fork could lift well over 2500 pounds at the maximum current of the motor, accounting for the 85-90% efficiency of ball screws. That was basically unnecessary, and should have been traded for speed somehow. Another factor was convenience – the clamp actuator and lift actuator were fundamentally the same parts, just one with a bigger gearbox.

There’s another detail which influenced the ability to grab and lift effectively. Look at where the forks end – well inside the tips of the triangular pontoons. They’re tucked in largely in the interest of anti-spinner protection – we didn’t want the forks to be the first thing a horizontal kinetic weapon hit. This also led us to make the “backboard” for the portion of the arm which dropped down into the dustpan.

It was kept that length also due to an unimplemented feature:

The crossbar at the front of the dustpan was going to be 1/4″ AR400 grade steel bars with some 1/4″-wall steel tubing and ribs backing it up. Not only would it have greatly increased the rigidity of the dustpan, but it would also have visually completed the bot, giving it more of a reason to occupy its nutty 42″ total length. One guess as to why we left if off…

By the way, after the tournament, I realized it was much better that it got left out. The triangular pontoons allowed OH1 to get a proverbial foot in the door in most of its  matches, where the broad surface of the crossbar might have prevented it, and also prevented more of the opponent from reaching the arm.

Can you guess the reason yet?!

In the end, besides “fundamental design flaws”, we never really had any major problems with the build or the tournament, and we frankly had a reasonable time with maintenance.  We got skilled to the point where we could pull a drive motor in under 5 minutes with tools ready, or the entire clamp arm assembly in the same time to get to the actuators, which was what we were doing when I was interviewed for the 2nd Lockjaw match; they had a penchant for bugging us for interviews and talks exactly when we needed to do something. It was definitely not “designed for service” in many ways, but we got good at it to make up for the part access deficiencies. It helped that OH1 was not very dense – a lot of it was air….

….but someow it STILL weighted 253 pounds out of 250 when done. Yup, no front crossbar for us.

So let’s summarize.

  • Overall, the completely unoptimized design, prioritized for a fast build and ‘unique combination’ of predecessor bots, led to a number of inherent flaws. When you glue 3 robots together in CAD, and have that be the point of the build, there will be parts of each bot which won’t work well with the others.
  • The lack of front traction under the dustpan section hurt maneuverability with an oppnent under control. Either I could try to push against someone or I could lift, but not both; at least, not very well.
  • The wheel were not pushed far enough towards the edge of the body to grant the bot sigificant rear traction. We tried to work with the existing spacing of the shuffler axles and used large wheels, both contributing to the short wheelbase.
  • The rubber shock mounts were used in bending instead of tension and compression like they were designed for. The result was an overly flexible dustpan that folded when the bot attempted a lift
  • The lifting forks were too slow because of the use of linear actuators. The lifting forks had superfluous force that could never have been used effectively.
  • The lifting forks did not project out of the dustpan, meaning the bot could not initiate a lift first. I could only drive under someone and try to maneuver them towards the arm.

Therefore, we can now designate a set of goals for Overhaul 2 that will resolve these problems.

  • Overall, the design must flow together and each part must have a reason to be in the place it is.
  • The chassis must either extend under the dustpan, or the dustpan must be affixed to the chassis in a way that allows a drive wheel to bear the lifting load; at least, the center of gravity should be between a drive wheel and a stationary contact with the ground, or best, between two drive wheels, as seen from the side.
  • The wheels must be made smaller to be pushed towards the corners, or the frame must have cut angles to permit tilting back; preferably 30 or 45 degrees or more. (Überclocker has wheels that ‘poke out the back’ for near universal traction)
  • The rubber shock mounts should be used in compression and tension as much as possible to mount the dustpan.
  • The lift system will be driven by a rotational actuator like a geared motor instead of a linear actuator; speed of lift is critical in being able to control opponent traction.
  • The lifting forks shall protrude from the dustpan profile, or better yet, be modular to be swappable to a short or long configuration as needed.

With these factors in mind, it was time to begin playing with concepts. We begin with the first concept of Overhaul 2, drawn some time in July 2015:

I’ll explain myself, don’t worry.