TRIMODAL — Build Guide¶

How to build a TRIMODAL-C from scratch. Consumer variant. No exotic materials. Target skill level: maker with electronics experience. Not an expert. Target cost: \(3,000–\)8,000 depending on sourcing and variant depth.

TRIMODAL-X build guide follows after TRIMODAL-C is community-validated.

Prerequisites¶

Skills¶

• Basic electronics: soldering, multimeter, oscilloscope
• 3D printing or CNC familiarity (for structural parts)
• Python programming (to configure and test the software stack)
• Basic mechanical assembly (nuts, bolts, bearings, actuators)

Tools¶

• 3D printer (250mm³ volume minimum) or CNC router
• Soldering station
• Multimeter + oscilloscope
• Torque wrench set (M3–M8)
• Heat gun
• Computer running Linux or macOS

Phase 0 — Software First¶

Before touching hardware, get the software running on your desk.

git clone https://github.com/jeanpaulniko/trimodal  # (planned)
cd trimodal
pip install -r requirements.txt
python tests/full_integration.py

Expected output: FULL INTEGRATION NOMINAL

If you see this, your development environment is correct. All 11 nodes simulated. You can now configure your variant before ordering parts.

Phase 1 — Structural Frame¶

Torso¶

• Material: Carbon fiber tube + Ti hardware (TRIMODAL-C)
• Geometry: Ellipsoid shell, 300mm × 200mm × 150mm
• Print/CNC: Upper shell, lower shell, midframe mounting plate
• Cargo bay: 180mm diameter × 280mm cylinder, press-fit into midframe
• STL files: hardware/cad/torso/ (planned)

Legs (×6 consumer, ×8 extended)¶

• 3 segments per leg: coxa (hip), femur (upper), tibia (lower)
• Joints: 2-DOF hip (pan + tilt), 1-DOF knee, 1-DOF ankle
• Each leg: 4 total actuators
• Material: PLA/PETG for initial build, upgrade to CF-PETG or aluminum
• Length: 200mm per segment, adjustable via mounting holes

Claws (×2)¶

• 2-DOF wrist + 3-finger gripper
• Retract flush against torso when not in use
• Silicone grip pads on finger tips
• Material: PLA + silicone + steel cable tendons

Phase 2 — Actuators¶

Each leg needs 4 actuators. Total: 24 (6-leg) or 32 (8-leg) + 6 claw actuators.

Recommended: Dynamixel XL430-W250 or equivalent¶

• Protocol: UART/TTL half-duplex
• Torque: 1.5 Nm (sufficient for 2.5kg per leg at 0.2m moment arm)
• Daisy-chain up to 253 per bus
• Drop-in swap for higher torque variants as budget allows

Alternative (lower cost): MG996R servo + custom driver board¶

• Requires position feedback loop in software
• HAL already supports both via SimActuatorDriver → swap to real driver

Wiring¶

• Per leg: single daisy-chain TTL bus to node SoC
• Power: 12V bus, separate from logic (7.4V) and signal (3.3V)
• Decoupling: 100μF per actuator, 10μF per logic rail

Phase 3 — Node SoCs (×11)¶

Each node is a sovereign compute unit. Start with Raspberry Pi CM4 or equivalent.

Consumer build: Raspberry Pi CM4 (4GB RAM, 32GB eMMC)¶

• Cost: ~$60/node × 11 = $660
• Connect via: SPI/I2C to sensors, UART to actuators, Ethernet to photonic bus sim
• Run: core/node.py + core/main.py per node
• Communication: UDP multicast simulating WDM (software-defined wavelengths)

Upgrade path: custom RISC-V SoC (when open hardware SiC boards available)¶

• Lattice iCE40 FPGA as interim quantum sim substrate
• SpacemiT K1 RISC-V for higher performance
• HAL swap: SimCommsDriver → EthernetCommsDriver → WDMPhotonicDriver

Node assignment¶

Node 0: torso     (hypervisor candidate, LLM inference primary)
Node 1-8: legs    (actuator control, local sensor fusion)
Node 9-10: claws  (manipulation, tool use)

Phase 4 — Sensors¶

Minimum sensor set per node:

Total sensor BOM: ~$400 for 11-node build.

Phase 5 — Power¶

Battery: 18650 Li-ion pack (consumer starter)¶

• 10S4P configuration: ~1480Wh
• BMS: ANT BMS 10S 100A
• Fits in cargo bay: 8.5L × 1600 Wh/L estimate
• Endurance: 1480Wh / 15W avg = 98hr (4.1 days)

Si-C upgrade (when available, 2027+):¶

• Same form factor, same BMS interface
• 600→800 Wh/kg → doubles endurance at same weight

Solar: flexible amorphous silicon panels¶

• Consumer: 0.4m² × 10% efficiency = 40W peak
• Mount: velcro-attach on torso upper shell (disposable/replaceable)
• Cost: ~$50 for the panel set

Charging: USB-C PD 65W (travel) + inductive pad (home base)¶

Phase 6 — Software Configuration¶

Once hardware assembled:

# 1. Flash each node
python tools/flash_node.py --node 0 --role torso
python tools/flash_node.py --node 1 --role leg_1
# ... etc

# 2. Calibrate IMUs
python tools/calibrate_imu.py --all-nodes

# 3. Zero actuators
python tools/zero_actuators.py

# 4. Run integration test
python tests/full_integration.py --hardware

# 5. First walk
python core/main.py --mode walk --speed 0.2

Phase 7 — RTSG Owner Binding¶

# Create owner I-vector profile
python tools/bind_owner.py --name "Your Name"
# Follow prompts: voice sample, face enrollment, gait calibration

# Set autonomy level
python tools/set_autonomy.py --level 3  # semi-autonomous default

# Connect to intelligence engine (optional but recommended)
python tools/connect_engine.py --url https://your-engine-instance/

Community¶

Build logs, issues, improvements: GitHub (planned) Wiki contributions: POST to smarthub.my/wiki/api/wiki/update Discord: planned

Share your build. Every iteration improves the design for everyone.