ZSR365

Thesis Project / Robotics

Unitree Go2 Voice Control System

A voice and text command prototype for a Unitree Go2 robot, designed around speech recognition, semantic parsing, command planning, safety checks, queueing, mock execution, and robot adapter layers.

Prototype PythonWhisperCommand PlanningSafety ControllerMock AdapterRobot Adapter
Abstract quadruped robot command interface visual

My responsibilities

  • Designed the command flow from user input to validated robot action.
  • Built the safety and command queue concepts for safer execution.
  • Separated mock execution from the future robot adapter.

Main features

  • Speech and text command input
  • Semantic intent parsing
  • Safety validation before command execution
  • Mock execution path for local testing
  • Adapter layer for future hardware integration

Outcomes

  • Created a testable control prototype before live robot deployment.
  • Clarified system boundaries between UI, planning, safety, and execution.

This project explores a safer control workflow for a robot dog command system. The first version focuses on the software architecture: accept speech or text input, convert it into a structured command, validate the action, then pass it through a queue to either a mock adapter or a future hardware adapter.

The main design goal is to keep the robot execution layer isolated from planning and safety logic. This makes it easier to test commands locally before connecting real hardware.

Related downloads