← All projects
FIRST Tech Challenge · DECODE · 2025–26

Scorpius

A vision-guided flywheel launcher on a custom carbon fiber tank-drive chassis. The turret tracks the goal automatically, and the wide intake lets the driver collect game pieces without precise lining up. More of the robot is custom-made than bought off the shelf.

Role
Design Lead · CAD + Build
Team
6832 Iron Reign (11 members)
Tools
Onshape, Fusion 360 CAM, Blender
Result
1st Place Design Award, 2026
Drag to rotate · Scroll or pinch to zoom
01 · Problem

The challenge

This season's robots collected game pieces called "artifacts" and launched them into a goal from one of two launch zones. Several constraints shaped our design:

  • Cycle time mattered most. Pattern bonuses only counted after each period, so fast, repeatable scoring was worth more than sorting.
  • The near zone was crowded and matches involved a lot of defense, so we wanted to score accurately from the far zone.
  • Artifacts were alliance-neutral, so both alliances fought over them. The intake had to be fast and forgiving.

Goal: move artifacts quickly and smoothly from pickup to launch, score consistently from anywhere on the field, and build a chassis tough enough for contested areas.

02 · Rapid prototyping

Launcher in 2 days

Right after the game was revealed, we built three flywheel launchers from MDF and REV rail in two days. We wanted to find out which approaches worked before committing to CAD.

  • Counter-rotating wheels: didn't reach enough height or distance, and shots were inconsistent because each wheel touched the artifact only briefly.
  • 3D-printed nylon flywheel wrapped in surgical tubing: reached enough distance, but the tubing stretched and left gaps, so shots were inaccurate.
  • Foam-filled tire at ~45°: the tread gripped well, but even at a 1:1 gear ratio it lacked speed.

What we learned: where and how the wheel contacts the artifact matters most. Contact far from the center turns wheel speed into spin instead of velocity. This confirmed that a custom flywheel would be the foundation of the robot.

03 · Iteration

Rimfire: the flywheel, five versions

Test data showed a large drop in RPM each time an artifact made contact. The flywheel wasn't storing enough rotational energy, so back-to-back shots fell short. Most of the redesigns below aimed to increase moment of inertia without making spin-up too slow.

VersionChangeOutcome
V1Lightweight carbon fiber frame, nubbed TPU tire, hub motor in a nylon shellSpun up quickly but vibrated heavily; low inertia and poor traction
V2–V3Rigid outer nylon shell, large curved-flap TPU tire, vented motor housingEnough launch power, but speed dropped after each shot and the single motor limited spin-up
V4Two external motors on belts, plus "momentum batteries": nylon pods holding 12 steel balls each (135.5 g per pod), with steel-filled PLA ringsAlmost no speed drop between shots and easier maintenance, but slower spin-up
V5Much smaller diameter, simplified 3-part design, steel flywheels, grip tapeHigher velocity and inertia, less backspin on artifacts, easiest to maintain
143 gRimfire V2 rotating mass
582 gV3.5 with momentum batteries (4× V2)
Quarticregression fit of distance to flywheel speed for auto-aim
04 · Subsystems

Intake, turret & chassis

Intake, 4 versions. V1 used a belt with TPU flaps and could only pick up artifacts from dead center. The V2 "wing" intakes were too complex and couldn't reach artifacts without exceeding the size limit. The V3 "converger" rollers funneled artifacts toward the middle, but the left roller interfered with the side belt. V4 removed the side belt and moved the belt into the center channel. This tripled the intake area, and its passive bobbing motion keeps artifacts from jamming.

Stinger launcher and turret. The launcher is a stand-alone module that slides in and out for maintenance. The turret is driven by herringbone gears to reduce backlash, and a Limelight camera tracks the goal's AprilTag. That means the driver doesn't have to turn the whole robot to aim.

Chassis. We chose an unusual two-wheel carbon fiber tank drive over a standard four-motor mecanum base. It freed up motors for other subsystems, and its shape allowed full and partial end-game parking despite its size. The tradeoff: no strafing, and the robot can pivot when hit away from the wheels. The plates use an 8 mm hole grid for future add-ons, every belt has tensioning slots, and the launcher has arced slots for adjusting launch angle.

05 · Manufacturing

Designed to be made in-house

We made nearly every custom part ourselves on our own CNC router and 3D printers. Making parts in-house saved roughly $2,000 and let us go from CAD to a test part in the same day. Milling carbon fiber was difficult at first; we wasted material by rushing until we learned the feeds, fixturing, and patience it requires.

Subsystem3D printed (unique / total)CNC milled (unique / total)
Chassis16 / 2915 / 23
Intake7 / 114 / 4
Stinger5 / 64 / 5
Rimfire7 / 104 / 4
Total35 / 5627 / 36

138 custom parts in total. About 269 hours of CAD and about 206 3D prints over the season.

06 · Results

How it performed

~10×increase in total points from Meet 1 to Regionals
9–12artifacts scored in the 30-second autonomous period
1stPlace Design Award, 2026

Each version shortened our cycle times. The turret's automatic aiming, flywheel speed set by distance, and one-button launch sequence took aiming and timing off the driver, so they could focus on strategy and defense.

What I'd change: the tank drive's lack of strafing and the robot's tendency to pivot when hit were real weaknesses in heavy-defense matches. I'd look at a compact swerve or hybrid drive next, using what we learned from our summer swerve project (below).

Summer project

Simply Swerve

Between seasons we develop our skills on open-ended projects. "Simply Swerve" is a swerve drive module and a three-module ("tri-swerve") chassis that we designed, tested at the Dallas Personal Robotics Group's Roborama competition (where it won a Da Vinci Award for innovative design), and published for other FTC teams to use.

Visualization

Reveal trailer

A reveal trailer for Scorpius, animated and rendered in Blender from the CAD model.

Next project

HawkBot · 2024–25