CADathon Team 235 Eclipse proudly presents Obelisk, our robot design submission for 2485’s Battery Battle CADathon. We were very happy to win the Robot Engineering & Industrial Design award. This robot and this design release was a collaboration between myself and @howlongismyname.
Main CAD Document
Technical Binder
2485 CADathon Awards Presentation
DESIGN/STRATEGY
Design Goals
In order to keep Obelisk relatively realistic to manufacture, we set a series of rules for our design. These are intended to encourage effective use of COTS parts and create a robot that an FRC team with mid-level manufacturing capabilities could reasonably build. We decided against using any pneumatics; with increased motor slots on the REV Power Distribution Hub and increased legality of linear actuators, it’s just not worth the weight for most teams to run pneumatics for just one or two mechanisms. We also decided not to use any COTS parts from Vex/IFI, in light of recent events.
235 Design Constraints:
- All custom parts must be able to be manufactured with the following equipment:
- Prusa i3 MK3S or equivalent FDM 3D printer with all-metal hotend and hardened nozzle
- Omio X8-2200 or equivalent CNC router
- Glowforge Plus or equivalent 40W laser cutter
- Band saw, chop saw, drill press, hand tools
- No pneumatics
- No more than 10 motors other than the drivebase
- No Vex/IFI components
Game Strategy
This CADathon game, Battery Battle, presents several unique challenges. While the main scoring element has a fairly low skill floor, orienting the Voltaic Piles correctly to score additional points in the Power Bank and the potential Ranking Point is very difficult without significant design complexity. While it is possible to simply score VPs without orienting them by just figuring out their orientation and placing them into the correct goals, this could mean missing out on a Ranking Point with bad luck.
We then decided the optimal strategy to complete a circuit for the RP would be to score VPs without orienting at first to save time, and then orient them correctly in the last few cycles when necessary. Because of this strategy, and also to reduce weight on our elevator and arm mechanisms, we made the decision to orient the VPs before the handoff to the manipulator arm.
These were our main goals, in no particular order:
- Score VPs in the Power Bank at all levels in correct polarity to score the Ranking point
- Drive up the ramp in endgame and use the grip feature to stay up
- Index and figure out game pieces to score inside of indexer to ensure no wasted cycle time while extended
- “Touch it, own it” full width intake with indexing to ensure no wasted cycle time
- Low COG to ensure easy driving
- Passive indexing
- Decrease DOF/complexity on elevator through rotation on indexer
This “Cardboard CAD” model was made on the first day of the CADathon as a very basic proof-of-concept for the layout of our mechanisms, and was used as a reference in the initial design stages. We did this before any of our master sketches, to get a better idea of how all the mechanisms would fit together in 3D.
ROBOT MECHANISMS
Drivebase
Before the kickoff of this CADathon, we agreed that we would aim to use REV MAXSwerve as our drivebase if possible, due to its incredible form factor. However, the Voltage Ramp and Conduit seemed to make 3” wheels almost impossible, so we decided instead to use the Swerve Drive Specialties MK4 swerve modules, mounted to the bottom of the frame to maximize ground clearance.
Drivebase Features:
- SDS MK4 in ground clearance configuration allows us to move quickly and precisely while still clearing the Voltage Ramp and Conduit
- L2 speed provides enough torque to drive up the Voltage Ramp long enough to deploy stinger
- Frame is ⅛” wall box tube, inner rails are Light MAXTube
- Electronics are stored underneath the frame with a removable panel for access
- Bumper mounting uses aluminum angle extrusion to support bumpers and toggle clamps to secure them to the frame
Intake
Once we decided that we would be intaking the Voltaic Piles in the horizontal orientation, a traditional over-bumper intake seemed perfect, especially with the 2022 season providing some great references. We based most of our design from the 2022 intakes of 3636, 1678, and 254.
The intake deploys with an asymmetrical four-bar linkage, using chain and sprockets on either side to equalize the torque between both sides. A NEO and MAXPlanetary powers the actuation, and another NEO powers the rollers, mounted on the arm instead of the main plates to reduce weight and keep the motor out of harm’s way.
The lower arm pivot points use the same ⅜” shoulder bolts as the tube rollers, with ⅜” ID oil-embedded bronze bushings.
Intake Features:
- Polycarbonate tube rollers with grip tape are lighter and grippier than traditional shaft and wheel rollers
- Dead-axle design with ⅜” shoulder bolts allows for easy removal
- Pivot points use ⅜” shoulder bolts and bronze oil-embedded bushings
- Rollers are powered by a NEO motor with a 2:1 reduction
- Intake deploy powered by a NEO motor with a 9:1 MAXPlanetary gearbox, 24:40 gear reduction, and 16:32 chain reduction for a total of 30:1 with fast extension and retraction for cycles
- Cross shaft equalizes torque on chain between both sides
Intake: Tube Roller System
This tube roller system is a design I’ve been developing for a while now. It’s originally based on 254’s intake rollers from 2022, which many other top teams designed their own versions of later in the season; the goal of mine is to create a very lightweight, easily serviceable roller design making use of shoulder bolts and 3D printed hubs. I designed the first iteration of it in April, after seeing many of these designs up close at Worlds. The current version is now part of the 235 CAD Library (more on that in the last section).
The rollers use modified 3D printed HTD pulleys as hubs on either end, using #10-32 screw-to-expand threaded inserts to fasten the tube, and 5/16”-18 threaded inserts on the ends for attaching the shoulder bolts. These could be 3D printed out of Markforged Onyx for maximum strength, but more accessible high-strength filaments like Prusament PCCF or Taulman Alloy 910 could work as well.
The lack of any metal hex shaft or tube stock anywhere on these rollers makes them very lightweight - they’re less than 70% the weight of an equivalent “traditional” intake roller using aluminum hex shaft and compliant wheels.
Because the shoulder bolt interfaces with the bearing instead of a shaft, to remove one of these rollers all you need to do is remove both shoulder bolts and remove it from between its mounting plates. This makes them incredibly easy to replace mid-competition. The shoulder bolt design also enables the use of ⅞” OD bearings, allowing for more material around the bearing hole, which can usually be a weak point of many intake designs.
The design is still evolving, and its weakest point right now is definitely the connection between the shoulder bolt and the hub pulley - relying on just half an inch of threads to secure the roller on each side is not ideal. I’m working on designing a new version using a longer shoulder bolt with a non-flanged threaded insert recessed further into the hub pulley, to increase the strength of the connection while maintaining the easy removability.
Indexer
Written by @howlongismyname
Wanting to make sure we didn’t have to put a 3 DOF system on our elevator, we chose to attempt to orient the Voltaic Piles in our indexer.
For this specific indexer, it was a very different design style from what I’m used to (box tubing + plate), primarily using parallel plates with nut strips for 90 degree joints. Additionally, the use of 0.09” aluminum plate is a first, as well as using the full REV ION ecosystem.
As the game pieces were relatively rigid, we trusted that they would sort themselves if you guided them in a funnel and used Omni wheels to agitate them.
Of course, the challenge is ensuring that you place the pieces in the correct orientation, so we decided with a “turntable” to flip the direction of the game piece. With a coaxial gearbox inspired by modern swerve modules and enabled by the REV MAXSpline bore gears, this allowed us to put a bearing and hex shaft through the gears, creating a COTS coaxial setup. We used a Lasershark LiDAR sensor and REV Color Sensor to determine the distance the game piece is from the “wall” of the indexer and the color, with the separate Omni wheels allowing for more precise control of the elements without having to integrate sensors or wires in a spinning system. Additionally, a Delrin ring is used to keep the whole turntable stable without adding too much complexity.
Overall, the goal was to create a somewhat simple setup that could be created with COTS parts that allowed for precise control and orientation of the pieces, allowing for quicker cycles while being mechanically robust.
To improve this setup, an absolute encoder setup with a gear with a REV through bore encoder attached to the main gear would allow for better control of the turntable, a third-party color sensor would allow for more accurate detection, as well as working on a setup that blocks external light to reduce inaccuracies.
Another option would be using a camera, which the driver or software could see the piece’s orientation, and orient that way. Additionally, a camera could eliminate the need to use a Lasershark for piece detection, although with machine vision complexity.
Indexer Features:
- Coaxial rotating spindexer for VP rotation with infinite rotation
- Uneven Omni wheel indexer to orient into spindexer
- Lasershark LiDAR sensor + REV color sensor for color detection and full hopper game piece detection
- MAXPlanetary gearboxes for packaging purposes and easy control of the parts.
- Wheeled indexer allows for easy programming and control
Elevator
Elevator, Arm, & Stinger CAD Document
With the elevator mechanism, we knew we wanted to take advantage of the incredible COTS parts on the market right now to make our design as effective as possible while staying within our manufacturing constraints. We decided on the West Coast Products clamping bearing blocks, as they enabled us to make our elevator frame using tube plugs instead of gussets.
The carriage ended up getting in the way of the shaft for the sprockets at the top. So, we decided to split the shaft into two smaller pieces, supported by small bearing plates attached to the top cross frame using nut strips. We were worried this could cause the two sides to get out of alignment and cause problems, so we added a cross shaft connecting the two with gears. However, upon further design review, this addition is definitely more trouble than it’s worth.
We also couldn’t fit traditional box tube bracing, due to the footprint of the indexer and intake in the front. So, inspired in part by 6800’s 2022 robot, we decided on carbon fiber tubes for our bracing instead. They attach using ball joint rod ends screwed into tapped aluminum standoffs that are epoxied into the ends of the tubes. This also makes the tubes effectively act as turnbuckles, keeping the elevator in tension. The tubes are 1 meter long, as that was the most easily available size within the range we needed, and we didn’t want to have to cut any tubes, to avoid dealing with the risk of carcinogenic carbon fiber dust.
Elevator Features:
- 3 Stage elevator with approximately 10’ of extension enables us to reach the highest bay of the Power Bank easily with plenty of stage overlap
- 2 NEO motors with 6.25:1 reduction extend the elevator to full height in approximately 2 seconds for quick cycle times while maintaining high stall load
- West Coast Products bearing blocks provide an elegant solution without the need for any complex in-house machining
- 20mm OD 1000mm long carbon fiber rods provide support and act as turnbuckles for structural tension
- MAXtube light grid pattern for ultra light elevator to keep COG low
Disclaimer: This elevator’s rigging is mostly CADathon BS. The first stage chain drive would work but pretty much everything above that is just random pulleys. Please do not attempt to build this in real life.
Carriage/Arm
This arm joint was mostly based on my pivot design for 3636’s 2022 climber, which worked very well throughout the 2022 season. (That design was itself made possible by a ton of advice I received from @Nick.kremer, big shoutout to him and 3512) The ⅞” OD aluminum round tube is nice and robust, and the ⅞” ID bronze oil-embedded bushings we used have a 1⅛” OD, meaning they fit perfectly into COTS sprockets with MAXSpline or standard 1⅛” bearing bores.
The carriage is packaged tightly thanks to a REV MAXPlanetary gearbox doing most of the arm reduction. The MAXPlanetary is side mounted to the cross tube of the carriage, using a .25” spacer plate and a .09” support plate on the other side to avoid crushing the 1/16” wall tube with through bolts.
Carriage/Arm Features:
- Arm pivot is a ⅞” OD aluminum round tube dead axle, using bronze oil-embedded bushings for rotation
- Arm rotation is powered by a NEO motor with a 75:1 MAXPlanetary and 16:48 chain reduction for a total of 225:1 reduction, which can rotate the arm 180 degrees in approximately one second
- 3D printed crush blocks in arm tube prevent it from deforming under the load of the arm pivot bolts
- Chain tensioner made of low-friction UHMW polyethylene tube
Roller Claw
This mechanism is overall pretty simple, with a few neat features. The roller claw holds 2 Voltaic Piles for maximum cycle efficiency, so that we can score both VPs in quick succession without doing multiple handoffs from the indexer. The claw picks up and releases the VPs easily using six pairs of 4” REV ION compliant wheels, all powered by a NEO with an 8:1 reduction and a gear reversal.
Now to the cool part: This “MAXHub pulley” I designed uses the same dead-axle ⅜” shoulder bolt and threaded insert system as the intake rollers, and uses the MAXSpline profile to interact with the REV ION compliant wheels directly without an additional hex hub. It has many of the same benefits as the intake rollers of being very lightweight and easy to remove.
Roller Claw Features:
- 4” REV ION compliant wheels pick up and release 2 Voltaic Piles easily
- Powered by a NEO motor with a total reduction of 8:1 with a compact gear reduction and reversal
- 3D printed pulleys with MAXSpline profile and threaded inserts create lightweight dead-axle rollers using the same ⅜” low-profile shoulder bolts as on the intake
Stinger
I’ll be honest, this mechanism was very much designed at the last minute. If we had the time to do it over again this design would have been much more robust and efficient, but as-is it’s at least a decent proof of concept. The robot drives up onto the Voltage Ramp, the stinger deploys down into the grip feature, and we’re done. @ArmourDuck did some really impressive physics calculations proving that the “yeet climb” would actually work, so with a better executed mechanism this would be a totally viable endgame solution.
Stinger Features:
- Deploys and locks into grip feature on Voltage Ramp to hold robot in place after driving up
- Powered by a NEO motor with 125:1 MAXPlanetary and 20:60 gear reduction
CAD PRACTICES
Multibody Modeling Workflow
Written mostly by @howlongismyname, edited by me
Team 235 utilizes multibody modeling in Onshape to speed up our processes and allow us to easily design in reference to other parts and speed up our assembly process. As a result, we primarily have one part studio and assembly for each subassembly. During our initial stages, we would do multiple subsystems per part studio. Still, it was challenging to have multiple people working on one part studio, and load times tended to be a detriment near the end of the CADathon. As a result, we chose to split some of the part studios into multiple part studios, particularly the intake and indexer subsystems. However, we wish we had split them into different documents to further help with load times.
A large emphasis is placed on modeling in reference to other parts, the assembly, and a top down approach using a master sketch, ensuring that parts are all relative to each other, and are accurate to the part studio, assembly, and sketches.
We also tend to spend a significant amount of the time initially in the part studio, only then transitioning to assembly once the part studio feels “complete” with all the necessary parts. As a result, the assembly process is swift, as everything is already figured out in the part studio.
Document Organization
Obelisk (aka 235-03) is organized into four documents: Main, Drivebase, Intake/Indexer, and Elevator/Arm/Stinger. In retrospect, the best change we could have made would have been to have a document for every major mechanism instead of combining them. By the end of this CADathon, Onshape was basically breaking constantly (over 10 million primitives!), and several last-minute design decisions had to be scrapped because we didn’t have time to wait several minutes for the document to reload every time we changed a single feature. Despite this, we did several other optimizations to our workflow that greatly improved things.
Composite Parts
A lot of our multi body modeling heavily relies on composite parts. For every new part studio we would derive various parts from other mechanisms for reference, and then combine them into a closed composite part. This has several benefits:
- Prevents duplicate parts when inserting to assemblies, as well as BOM tracking
- Allows for designing in reference to master parts
- Makes it very easy to distinguish derived parts from the rest of the part studio, especially by changing the color of the composite part
- Can show or hide all derived parts at once easily
Master Sketch
Our master sketch focused primarily on the positions of each mechanism relative to each other, with general space allocations for each mechanism rather than fully detailed sketches. We were able to figure out how much space each mechanism would have, and then use that to inform our subsystem documents by deriving the master sketch into each main Part Studio. We updated the master sketch and versioned the main document occasionally if we realized that a mechanism needed more or less space than we thought, but for the most part our initial expectations turned out to be correct.
Once the master sketch was done, additional subsystem layout sketches in the subsystem part studio would be made to plan the subsystem, in which the parts would reflect the layout sketches and the master sketch.
Relative & Accurate Part Studios
A significant emphasis is placed on ensuring that our part studios are in the absolute position relative to the exact origin of the assembly. As a result, the part studio closely reflects the assembly, with derived parts such as bearings or gears inserted for clearance or tolerance purposes. Additionally, this allows for a truly parametric approach to shafts with the Shaft Generator featurescript, allowing the part studio to automatically adjust shaft lengths when a sketch or part is modified. This approach also helps us put together our assemblies very quickly, as a few Group mates can constrain almost all of an assembly immediately.
While we hope to model and reflect the assembly as closely as possible, there is also an emphasis on ensuring that there are no duplicate parts when inserted into the assembly for cleanliness’ sake. Still, if necessary, duplicate parts are approached with the same composite part feature as mentioned above, turning one of the parts into a composite part.
Folders & Feature Naming
Naming all features in a Part Studio, while a good practice, is a bit of a pain to do, especially for a robot as detailed as this. Instead for the most part we just grouped features into folders named for what specific part or section they were for, and named the ones that we were editing the most. We did name almost all of our sketches, especially larger layout sketches of mechanisms, since we went back to those quite often. Overall we found this to be a pretty effective naming system, as it’s a nice balance between naming absolutely everything (which takes forever) and having no names and folders at all (which makes it nearly impossible to know what anything is).
Part Studio to Assembly Practices
- Group Mate
- Allowed for a quick initial assembly of the imported part studio, saving a chunk of time
- Replicate
- Used primarily on fasteners to model quickly in a part studio. Additionally, the use of a slightly different hole size can differentiate between rivet and bolt holes to make replicating easier. (For example, we used .196” for bolt holes and .1961” for rivet holes)
- Assembly Configurations
- We used configurations in all of our major assemblies to show different states of the robot (starting, intaking, scoring, etc.). This was done by setting limits on different mates in the assembly.
Part Studio Screenshots
Typically, all the sketches are hidden while working in the part studio, but they are shown for reference purposes. Note that the part studio closely reflects the assembly, and the use of light colored composite parts to differentiate them as reference parts.
235-03-0002 Intake/Indexer Document
Something to note in this document is how close the subassemblies are, and the use of deriving and modeling multiple plates to reference clearance. Initially a merged part studio of both intake and indexer, they were split to make the process of working on the subsystems easier.
- Intake Part Studio
- Indexer Part Studio
253-03-0003 Elevator Document
The layout sketches for the elevator are on the front plane, and the structure is extruded to two points from the side profile sketch. This keeps the sketches simple without the need for additional offset planes.
As the elevator is somewhat isolated, there wasn’t as much of a need to composite and derive the drivetrainn, and the mounting plates were modeled using the hole pattern in a sketch.
- Elevator Part Studio
- Arm/Roller Claw Part Studio
How this helps our document organization
As mentioned earlier, by condensing our part studios and assemblies, and properly referencing the sketches and parts via the derived feature, it makes it much easier to navigate and work in the various documents. For example, if someone wanted to work on the intake, it would be easy to find the intake part studio, and intake assemblies, and go straight to working on it without having to jump between multiple part studios or assemblies. Once complete, they can version to main and update it, without worrying about the load times too much.
The style of organization helps with our load times significantly. The main assembly only has to load a saved version of each document, as opposed to updating in real time; and each subsystem document only has to load its own features. Considering how hard we pushed Onshape with four documents, doing this robot in a single document would have been nearly impossible.
Utilizing the best possible tessellation option, there is quite the amount of detail, though it’s closer to 20 million with automatic tessellation. Compared to previous robots that either of us have designed, this has 2-4x more detail, yet is still reasonably workable due to how we split our document structure.
CAD Library
During and after this CADathon, Team 235 has been building up a small collection of useful parts that we tend to use often. This includes many configurable parts, as well as some assorted COTS parts that aren’t in MKCAD.
Some highlights are:
- Configurable ⅜” shoulder bolt with options for low-profile and ultra-low-profile
- Configurable bolts and rivets
- Two commonly used bronze bushings (⅜” ID for shoulder bolt pivots, ⅞” ID for arm joints)
- Several configurable HTD pulleys with various features
- WCP 12t HTD pulleys with 8mm keyed bore (still not in MKCAD, even though they were requested almost a year ago)
- Configurable round tube spacers
- Configurable Tube Roller system
- Sets of four swerve modules configurable by frame size
