Felix Hu

Felix Hu

Explore the future alongside industry practitioners and enthusiasts like you

ROBOCON Multi-Robot System

ROBOCON Multi-Robot System for Retrieval and Targeting Challenge

ROBOCON 2023 | 2022-2023 | Wuhan, China

Team: Jiacheng Li Yongye Hu Zhixiong Zhang Kehao Chen Guanshao Lin Junbo Shan Minkang Gong Chao Liu Xingzhi Li Yeying Wang Show all

Advised by: Professor Yu Hou

Integrated Mechanical Engineer

Reached the Top-16 finalist stage at ROBOCON 2023 out of 80 competing teams.

  • CAD
  • CNC machining
  • STM32F407
  • CAN bus
  • Pneumatics
Case Type Multi-robot competition system
Technical Center Mechanical integration, Elephant chassis control, field reliability
Outcome Reached the Top-16 finalist stage at ROBOCON 2023 out of 80 competing teams.
Ownership Owned mechanical design across the program; solo-authored Elephant chassis control; collaborated on other firmware and strategy.

Hardest engineering problem: The system had to stay competition-reliable while CAD, CNC fabrication, pneumatics, embedded controls, and field failures were all changing under deadline pressure.

Overview

This ROBOCON system combined three competition robots for object retrieval, launching, and speed-focused traversal. The work sat at the boundary between mechanical design, manufacturing, embedded controls, and field reliability under a fixed competition schedule.

My Role

I owned the mechanical side across all three robots: CAD, CNC drawings, BOM planning, supplier coordination, assembly, and failure-driven redesign. On firmware, I wrote the Elephant 3-PID dual-mode chassis controller solo; the remaining firmware was mixed-authorship with teammates, so I describe that part as collaboration rather than sole ownership.

Key Results

  • Reached the Top-16 finalist stage at ROBOCON 2023 from a field of 80 teams.
  • Selected a 3-wheel swerve architecture for Elephant, prioritizing top-end linear speed and gaining 33% over the prior chassis.
  • Built a competition-ready mechanical system across three robots using CAD, CNC machining, FDM printing, pneumatics, and iterative assembly testing.
  • Improved public-match reliability after repeated development failures, including fastening changes that addressed thread-stripping under impact loads.

3D Model Overview

System Architecture

ROBOCON multi-robot architecture and ownership diagram
The public system view keeps the three robots together while making ownership boundaries explicit: mechanical integration across the program, solo Elephant chassis firmware, and mixed-authorship firmware elsewhere.
ROBOCON responsibility map separating owned, collaborative, and separate-case work
The case is broad, so the responsibility map is part of the technical story: it keeps mechanical ownership, solo controller work, collaborative firmware, and the Dog deep dive distinct.

Elephant

High-speed 3-wheel swerve chassis with steering, drive, slip-ring packaging, and a dual-mode chassis controller.

Rabbit

Retrieval and launch robot with STM32F407 modules, motor control, pneumatic actions, and teammate-shared firmware.

Dog

Quadruped obstacle robot, now described separately as its own deeper embedded-control case.

Software / Firmware Architecture

SubsystemArchitectureOwnership boundary
Elephant chassisThree PID loops for X, Y, and yaw, with velocity/position modes and yaw compensation.Solo-authored chassis controller.
Rabbit actionsSTM32F407 firmware integrating CAN motor control, UART, launcher, lift, and clip actions.Collaborative firmware work with teammates.
Team strategyRobot roles, field timing, and match behavior were tuned as a multi-robot system.Shared strategy and field debugging.

Control / Simulation Methodology

The Elephant wheel architecture was the load-bearing design choice. A 3-wheel swerve drive added complexity through steering motors, drive motors, slip rings, belt steering, and continuous rotation packaging, but the speed gain mattered more than the extra build burden for the competition objective.

The controller design matched that mechanical choice: PID loops handled translational and yaw error while inverse kinematics and yaw compensation converted field-level intent into chassis motion. The portfolio should show this as a practical embedded-control architecture, not as a black-box robot.

ROBOCON Elephant chassis controller from field commands through PID, yaw compensation, inverse kinematics, and swerve modules
The Elephant controller diagram narrows the broad ROBOCON case to Felix's strongest firmware slice: field command, mode selection, X/Y/yaw PID, yaw compensation, swerve mapping, and wheel-module feedback.

Rabbit had a different control problem: it needed to coordinate chassis movement with ring capture, lift/reload, and pneumatic launch actions. The firmware separated joystick/action dispatch from chassis kinematics and mechanism commands, so the team could tune drive behavior, clip force, lift increments, and launch timing without turning the robot into one monolithic routine.

ROBOCON Rabbit control and action pipeline from SBUS input through chassis kinematics, CAN motors, lift, clip, and pneumatic launch
The Rabbit method diagram shows the collaborative firmware slice: SBUS input fed remote dispatch, chassis commands became yaw-compensated four-wheel motor speeds, and mechanism actions stayed separated as clip, lift, and launch routines.
WorkstreamHow to read the claimEvidence direction
Mechanical programFelix-owned CAD, CNC drawings, BOM planning, sourcing, assembly, and failure-driven redesign across the three robots.Use this as the broad system-integration claim.
Elephant chassis controllerSolo-authored 3-PID chassis controller with yaw compensation and inverse-kinematics conversion.Use as the strongest firmware slice on this page.
Rabbit and shared firmwareCollaborative STM32F407 firmware combining SBUS dispatch, 4-wheel chassis kinematics, CAN motor control, clip/lift routines, and pneumatic launch relays.Use as a concrete method slice, while keeping authorship collaborative.
Dog quadrupedSeparate deep embedded-control case.Link readers to the Dog page for details.

Validation & Debugging

ROBOCON field reliability loop from CAD and CNC through assembly, field failure, redesign, retest, and match reliability
The reliability loop shows why this is a system-integration case: CAD, CNC drawings, assembly, impact failures, fastening redesign, pneumatic timing, and controller tuning all fed public-match reliability.

Validation happened through field iteration, public-match reliability checks, and repeated mechanical failures. The most useful debugging pattern was not one isolated algorithm fix, but the loop from CAD to CNC drawing to assembly to impact failure to fastening redesign.

The system result was Top-16 at ROBOCON 2023 from 80 teams. That result anchors the mechanical and firmware work without overstating sole ownership of every subsystem.

What I Would Improve

In hindsight, I would split the page into a system overview plus linked robot-specific deep dives only after the Elephant and Rabbit casefiles are equally mature. I would also take more solo firmware ownership earlier, because the mechanical system was already strong and the next growth area was embedded-control fluency.