Dash

An 18-DoF open-source humanoid robot developed from scratch — custom motor controller design, firmware, communication architecture, and RL policy deployment.

Overview

Dash is an open-source humanoid robot platform developed independently from the ground up, spanning custom actuator electronics through learned locomotion control deployed on hardware. The system integrates a custom-designed motor controller and communication architecture, firmware, and a reinforcement-learning-based control stack into a single, fully open humanoid platform.

Dash, an 18-DoF open-source humanoid robot.

Specifications

   
Height 110 cm
Weight 35 kg
Degrees of freedom 18 total — 5 DoF per leg, 4 DoF per arm
Actuation Cross-roller bearing motors at the shoulder pitch, hip, and knee joints for high dynamic motion and heavy payload handling
Electronics Fully open-source design, including custom motor controller PCB and firmware

Motor Controller & Communication Board

The motor controller and communication board form the electrical backbone of the robot, providing closed-loop actuator control and a communication bus linking every joint to the central compute.

Left: custom motor controller board. Right: communication board.

Motor Controller Specifications

Designed to an ODrive Pro-level specification:

   
Voltage input 12–54 V (tested)
Control loop frequency 40 kHz
PWM switching frequency 40 kHz
Communication CAN 2.0
Position sensing 14-bit onboard encoder, with external encoder support via SPI
Thermal protection NTC thermistor sensing

Firmware

The firmware running on each motor controller handles low-level commutation, current sensing, and closed-loop torque and position control, and exposes the CAN interface used for coordination across the robot.

Motor Testing

Each motor and controller pair was bench-tested in isolation to validate control performance before integration onto the robot.

Bench testing a motor with the custom controller and firmware.

Robot Integration

With all motors validated individually, they were integrated onto the robot frame and brought up as a complete system.

Integration testing of all motors on the assembled robot.

RL Policy Deployment

A reinforcement-learning locomotion policy was deployed onto the physical robot, closing the loop from custom electronics to learned control.

RL locomotion policy deployed on the physical Dash robot.