Skip to content

LLM-Controlled Robotic Dog

Robotics

Robotic dog controlled through natural language. Claude API + ESP32. Tell it to do backflips, push-ups, or choreographed sequences, and it plans the rest.

Robotics
Claude
ESP32
Python
Microcontroller
LLM
Hardware
Computer Vision

Key Features

  • Natural-language command interpretation with the Claude API
  • Serial communication with the ESP32 microcontroller
  • Autonomous multi-step action sequences
  • Gyroscope-based auto-correction for balance
  • Complex choreography engine (backflips, push-ups, stretches)
  • Real-time movement coordination
  • Hardware-software integration with LLM control
  • Conversational robotics interface

This is a robotic dog that you control when you talk to it. You tell it “do three backflips then take a bow”. Claude converts that sentence into a movement sequence, and the ESP32 executes the sequence. There is no pre-programmed command list.

How It Works

The system has three parts. A Python host sends your command to the Claude API. The Claude API handles the understanding: context, intent, and the sequence of movement primitives. An ESP32 handles the execution: servo control with gyroscope feedback in a closed loop. As a result, the robot corrects its balance mid-maneuver. A serial protocol carries commands between the host and the ESP32.

What It Can Do

  • Backflips with landing stabilization
  • Push-up sequences
  • Stretches
  • Custom action chains built from your commands

The IMU auto-correction makes the backflips land. Without this correction, the robot lands on its back.

Demo

The video shows how the robot interprets commands and chains moves: youtu.be/Xs0VmfZI658