Neurohelmet

From Thought to Machine — DIY Brain-Computer Interface

Overview

A four-phase project to build a brain-computer interface from scratch, use it to control a BattleTech-inspired mech game, and then pilot real robots through the robonet mesh.

This isn't a consumer product. It's a father-son engineering project that starts with a soldering iron and ends with a robot that moves when you think at it.

Phase 1 — Hardware

The Cerelog ESP-EEG is an open-source 8-channel EEG board built around the ADS1299 (24-bit ADC) and ESP32-S3 (WiFi + BLE). The QFP-64 package requires hot air reflow — not beginner soldering, but achievable with a $50 station and patience.

Target: a working BCI board for under $100. Commercial alternatives (OpenBCI, Muse) start at $350+ and don't teach you anything about the hardware.

Eight electrodes on the scalp. Motor cortex positions (C3, C4, Cz). Dry electrodes for daily use without conductive gel. 3D-printed headband or modified helmet shell.

Phase 2 — Signal Processing

Raw EEG is noise. The signal processing pipeline turns it into intent:

• Bandpass filter (1-40Hz, notch at 60Hz to kill power line noise)
• Artifact rejection (eye blinks, muscle tension, electrode pops)
• Feature extraction (mu rhythm 8-12Hz, beta 13-30Hz — the motor imagery frequencies)
• Classification (left/right motor imagery, concentration, relaxation)

Think LEFT — left hemisphere mu rhythm desynchronizes. Think RIGHT — right hemisphere. Concentrate — frontal beta increases. Relax — global alpha rises.

Four control signals from thought alone.

Phase 3 — The Mech Game

BattleTech neurohelmet as game mechanic. The pilot's concentration maintains mech balance. Lose focus and the mech staggers.

• Balance: continuous concentration keeps the mech upright
• Movement: motor imagery controls turning and throttle
• Targeting: head orientation or gaze direction
• Weapons: button hybrid (BCI for movement, hands for weapons)

Multiplayer arena combat. Better neurohelmet skill = better pilot. The brain IS the controller.

RoboRally mode: program a movement sequence by thinking LEFT/RIGHT/FORWARD in order. All robots execute simultaneously. Physical robots on a real grid mirror the game via robonet mesh.

Phase 4 — Robot Pilot

The same BCI that controls the virtual mech now controls a real robot. Differential drive as MVP — two wheels, tank steering mapped to left/right motor imagery.

Commands route through the robonet mesh. Any node can pilot any robot. Latency-aware routing. FPV video from robot camera to pilot display via WebRTC.

The game trains the neural pathways. The robot makes them real.

Why Build This

Because a kid asked about Neuralink and the answer is "we can build one at home." Not the invasive surgery kind. The kind where you solder a board, 3D-print a headband, write some Python, and drive a robot with your brain.

The barrier to BCI is not technology. It's the belief that it requires a billion-dollar company. It doesn't. It requires an ADS1299, an ESP32, and a willingness to learn.