ace-trump-tech/MindPaw

What it solves

MindPaw provides an affordable, open-source platform for building a desktop quadruped robot that integrates AI interaction, computer vision, and emotional response. It lowers the barrier to entry for learning robotics by combining hardware design, firmware, and AI integration into a single project with a total material cost of approximately ¥50.

How it works

The robot is powered by an ESP8266 microcontroller and utilizes several modules for interaction:

• AI Interaction: Connects to the Doubao LLM via API for intelligent conversations, which are translated into physical movements and facial expressions.
• Perception: Uses an OV2640 camera for gesture recognition (via a lightweight MLP) and a v2.0 streaming 3D reconstruction layer for obstacle avoidance and point-cloud visualization.
• Interaction: Employs an HLK-V20 voice module for command recognition, an OLED screen for displaying 7 different emotions, and a speaker for audio feedback.
• Control: Can be operated via a web-based remote control interface hosted on the device.
• AI Infra Gateway: An optional Python-based gateway that handles model adaptation, authentication, and structured output for researchers studying AI agents.

Who it’s for

• Beginners and Hardware Hobbyists: People wanting to build their first robot and learn about ESP8266, servos, and basic electronics.
• AI and Edge Computing Researchers: Developers interested in AI agents, model routing, and the coordination between edge devices and cloud LLMs.

Highlights

• Ultra-Low Cost: Complete build costs around ¥50 (basic version ¥35).
• Multimodal Interaction: Supports voice commands, gesture recognition, and AI-driven dialogue.
• Emotional Engine: Features a state machine based on the PAD model to simulate moods like happiness or boredom.
• Full Open Source: Includes all C++ firmware, PCB schematics (LCSC EDA), and 3D printable STL files.
• Streaming 3D Perception: v2.0 adds real-time 3D reconstruction for obstacle avoidance with ~150ms latency.