Miles’s sumo robot journey

New build · August 2026

Version 2: built for CMRG 2026

Version 2 is not just a new shell. It is the result of testing the radio, learning the competition rules, choosing compatible parts, measuring everything for CAD, and changing the design when the evidence showed that the first hardware was not strong enough. Chiang Mai Robot Games runs from 28–30 August 2026, so this is our competition-weekend build update.

What I did on the way

1. Reviewed the official rules: confirmed the 1 kg Junior class, 20 × 20 cm starting limit, wireless control and the requirement for a programmable control board.
2. Proved the control idea: tested the HotRC transmitter and receiver, read steering and throttle pulses, and programmed tank-style mixing with a safe stop when the signal is lost.
3. Upgraded the drivetrain: moved from the learning setup’s L298N and TT motors to a Cytron MDD10A and two matched 12 V, 580 RPM GM25 metal gearmotors.
4. Selected the power system: planned around a HeliCoxNano 2S 7.4 V, 1100 mAh, 30C LiPo and an LM2596 converter for regulated 5 V logic power.
5. Designed before printing: measured component envelopes, checked the 200 × 200 mm layout in Tinkercad, left room for wires and terminals, and kept the battery low and removable.
6. Built the competition body: assembled a compact black chassis, installed the electronics and added a low yellow front wedge.
7. Kept a real engineering record: saved test programs, wiring diagrams, parts decisions, safety notes and CAD reviews instead of hiding the mistakes and changes.

The starting point

Version 1: what the first robot taught me

Why document a robot that is not perfect yet? Because the failed paths and decisive tests are the useful part. This page separates what we proved from what we still suspect.

Build photos

What we built

HotRC transmitter  ~~ radio ~~>  F-06A receiver
                                      │ CH1 + CH2 servo pulses
                                      ▼
                                  Arduino Uno
                                      │ D5–D8 direction signals
                                      ▼
4 × AA motor pack  ──────────────>  L298N driver  ─────>  2 × TT motors
       motor power                     ▲
                                       │ shared GND

Laptop USB (bench testing) ─────────> Arduino + receiver logic power

The Uno reads radio pulses and makes driving decisions. The L298N is the muscle: it switches the separate motor supply through two H-bridges. The Uno must never power the motors directly.

Confirmed specifications

Verified wiring map

Disconnect the motor battery before moving any wire.

L298N header:  ENA | IN1 | IN2 | IN3 | IN4 | ENB
                     jumper  D5    D6    D7    D8   jumper

Motor wires belong on OUT1–OUT4, never on Arduino pins. Reversing the two wires of one motor safely reverses its physical direction. ENA and ENB remain jumpered for full speed.

Power lessons we learned

• Arduino GPIO pins provide instructions, not motor current.
• The L298N loses voltage internally, so motor-terminal voltage is lower than battery voltage.
• LEDs can stay illuminated while weak cells collapse when both motors start.
• Solder and insulate motor leads; loose hook connections caused intermittent behavior.

Diagnostic timeline

1. USB discovery: macOS detected the Uno clone as CH340 USB Serial at /dev/cu.usbserial-1110.
2. Compilation: early errors came from pasting fragments inside loop(). A complete sketch with setup(), loop(), and motor helpers compiled successfully.
3. Receiver identification: the F-06A outputs servo-style PWM, not separate forward/back/left/right digital signals.
4. Six-step motor test: IN1, IN2, IN3, IN4, both A directions, and both B directions all produced movement—six of six.
5. Radio capture: more than 1,100 readings showed stable CH1 and CH2 pulses and revealed that the original 1400/1600 thresholds did not match the real endpoints.
6. Control-model correction: a pistol-grip radio needs steering/throttle differential mixing, not one receiver channel assigned directly to each motor.
7. Remaining symptom: receiver-driven tests became intermittent, particularly on the right motor, even though the automatic output test had proved that channel could operate.

Where we would value community help

The evidence points toward an intermittent connection, motor-supply sag, or a timing/ failsafe integration issue—not a completely dead motor or H-bridge. If you have built a similar Uno + F-06A + L298N robot, these are the most useful next checks:

• Measure AA-pack and L298N output voltage while both motors are commanded.
• Continuity-test D7→IN3, D8→IN4, ENB, OUT3 and OUT4 while gently moving each connector.
• Confirm whether sequential pulseIn() reads are reliable enough for both simultaneous F-06A channels, or replace them with interrupt-based capture.
• Recommend a more efficient dual brushed-motor driver for 3–6 V TT motors, if locally available.
• Suggest safe differential-mixing and failsafe values for CH1 1230–1729 µs and CH2 1030–1877 µs.

Please share advice with a reproducible test: what to disconnect, what to measure, the expected result, and what each possible result means. That style of help moved this build forward far more than swapping several things at once.

How these Version 1 priorities shaped Version 2

Credits and links

• Builder: Miles Michelini — competition story
• Family and build support: Mike, Wendy and LaoYe
• Event: EasyKids Robotics Competition 2026, Chiang Mai · August 2, 2026
• Version 2 event: Chiang Mai Robot Games 2026 at Yupparaj Wittayalai School · August 28–30, 2026
• Purpose: Share the real process so another young builder can diagnose one variable at a time.