Biped Robot Control System

An ESP32-based six-servo biped robot with:

• a desktop control app for calibration, pose editing, and operator workflows
• a Streamlit dashboard for advanced diagnostics and experimentation
• color-based computer vision for robot-and-ball tracking
• autonomous forward movement until ball collision

Highlights

• Automatic device discovery on the local network
• Flash-backed pose storage and sequence playback
• Per-joint calibration and physical-angle mapping
• Live annotated camera feed from Android IP Webcam
• Yellow-marker robot detection and red-ball detection
• Video recording and export for autonomous runs

Robot Mapping

The firmware keeps legacy HTTP joint ids, but they map to the biped legs as follows:

API name	Physical joint
base	Left Ankle
shoulder	Left Knee
elbow	Left Hip
wrist_pitch	Right Knee
wrist_rotate	Right Ankle
gripper	Right Hip

All six servos are configured as positional_180.

Current stand pose:

Joint	Angle
Left Ankle	80
Left Knee	90
Left Hip	110
Right Knee	25
Right Ankle	40
Right Hip	135

Main Interfaces

Desktop app

Primary operator app:

• biped_desktop_app.py
• launched with biped_desktop_app.bat

Main capabilities:

• connection and automatic discovery
• flash save and reload
• pose editing and sequence tuning
• per-joint control and calibration
• Wi-Fi and controller configuration
• live vision feed
• autonomous forward-until-collision mode

Streamlit dashboard

Advanced dashboard:

• streamlit_app.py
• launched with biped_streamlit_dashboard.bat

Useful for:

• deeper diagnostics
• sequence recorder workflows
• gait experiments
• lower-level HTTP visibility

Default Motion Workflow

The current forward motion sequence is:

stand,left_forward,right_forward,stand

This sequence is tuned on the real robot and used by the vision workflow as the default forward action.

Vision Workflow

Vision modules:

• biped_vision_camera.py
• biped_vision_tracking.py

Behavior:

• detect the yellow robot marker
• detect the red ball
• estimate robot-to-ball distance from image geometry
• keep moving forward while both are visible and distance is above the collision threshold
• stop automatically when the collision threshold is reached

Project Structure

• biped_robot.ino — ESP32 firmware
• biped_desktop_app.py — primary desktop operator app
• streamlit_app.py — advanced dashboard
• biped_endpoint_discovery.py — shared network discovery
• biped_gait_phase.py — host-side gait helpers
• output/ — poster, screenshots, and demo video
• REPORT.md — report source
• INSTALLATION.md — setup guide

Screenshots

Setup

Joint Calibration

Wi-Fi

Controller

Vision

State

Deliverables

• Poster: output/poster.jpeg
• Demo video: output/vision_recordings/autonomy_20260513_232641.mp4
• Report source: REPORT.md

Setup

See INSTALLATION.md.

License

This project is released under the MIT License.