The team
285C, Discobots Chaos, is a VEX V5 team from Carnegie Vanguard High School in Houston. I've done VEX for three seasons. In the 2025-26 Push Back season I led programming and was also on the build and drive teams.
We qualified for the 2026 VEX Robotics World Championship through a skills rolldown after the Texas Region 3 Championship. It was the first time a CVHS team had qualified in over five years, and all three 285 sister teams qualified together.
The season
| Event | Date | Rank | Furthest round |
|---|---|---|---|
| Katy Holiday Classic | Nov 2025 | 34th | Did not qualify |
| Den Wars | Feb 2026 | 4th | Semifinals |
| New Year New Drive (GORTS) | Feb 2026 | 22nd | Semifinals |
| Texas Region 3 Championship | Feb 2026 | 6th | Quarterfinals |
Our combined skills score went from 100 to 102 to 137 points across the last three events. At States we went 7 and 2 in qualifications and scored 80 autonomous points, the most of any team at the event.
The robot
A six-motor, six-wheel tank drive geared 3:4, so 600 RPM motors give 450 RPM at the wheels, with 3.25-inch omni wheels on the outside and traction wheels in the middle for lateral stability. A two-motor intake on a dual chain feeds a conveyor that can score in the long goal or, with a piston pulling the conveyor floor down, the middle goal. Pneumatics run a scraper for match loads and a wing for descoring. After a collision at Den Wars bent the frame and stopped one side from moving, we added cross-bracing; after Regionals we angled the front of the drivetrain so blocks get guided into the intake instead of deflected.
The code
Everything runs on PROS with LemLib for odometry and PID. Position comes from the IMU plus a vertical tracking wheel on a rotation sensor. Both PID loops use only P and D terms, since nothing caused the kind of persistent offset that integral correction is for. Lateral slew is on because the robot's center of gravity is high enough to tip if it accelerates too hard. Driver control uses a steering curve with a small deadband and an exponential gain, so slow movements are fine-grained but full speed is still there. The brain screen runs an LVGL autonomous selector over seven match routines and one skills routine.
Motion chaining
The part I'm proudest of. LemLib can chain one motion into the next without stopping, but each motion needs two hand-tuned values: how early to exit, and what speed to hold on the way out. Tuning those for every motion in every routine was slow and fragile, so I wrote a CustomChassis class that derives both from the distance of each motion.
Fig. 2. Same path, two ways to drive it
This simulation needs JavaScript. The idea: instead of stopping at every waypoint, the robot leaves each motion early at half its peak speed and carries that speed into the next one.
Here d is the distance of the motion and k is a fixed fraction, around 15 to 20 percent for drives. Turns work the same way with angular distance and a tighter window, since overshooting a heading throws off everything after it. The full reasoning is in this post.
The template
Every season starts with the same boilerplate: lifecycle callbacks, drivetrain and odometry setup, an autonomous selector. I pulled ours out into a public template with the code split into small single-purpose files, so we can start the next game from a working skeleton instead of rewriting it.