MindPaw
🐕 MindPaw — 基于 ESP8266 的桌面级四足机器狗,支持语音交互 (HLK-V20)、手势识别 (OV2640+轻量MLP)、AI 对话 (豆包大模型)、PAD 情感计算、WiFi 网页遥控。总物料成本 ≈¥50,代码/PCB/3D打印全开源,适合机器人入门、嵌入式AI、HRI 情感交互研究。
MindPaw is an open source desktop robot dog
MindPaw is a desktop quadruped robot built on an ESP8266 with voice, gesture and AI chat features, designed to cost about 50 yuan in parts.
What MindPaw is
MindPaw is a desktop quadruped robot project based on the ESP8266 microcontroller. The README says its distinctive trait is that it is not just a remote controlled car: you can talk to it, gesture at it, and it reacts with emotions. It connects to the Doubao large model for intelligent conversation, and the README claims the whole bill of materials comes to under 50 yuan, with a basic version at about 35 yuan. Code, circuit diagrams and 3D printable files are all provided. The author states the project was built from scratch rather than from a ready made development board kit, so the code and documents are meant to help someone learn how to combine an ESP8266, servos, a camera, a voice module and AI. The README frames two paths through the repository: a demo path for beginners who want to reproduce and play with the dog, and an AI infrastructure path for researchers in edge computing and agents. Both share the same firmware and hardware, and a beginner can leave the gateway address empty to stay in direct connect mode. The project is written for makers who want a cheap, open entry point into embodied AI rather than a polished product.
Features and hardware list
The feature table in the README maps triggers to behaviors. Voice control runs commands like forward, happy or hello, and switches expressions or plays sounds. Gesture control uses a camera to recognize waves, fists or pointing and then acts. Web control opens a full remote interface at the local address 192.168.4.1 from a phone browser. AI conversation calls the Doubao API and replies with motion and expression. An AI infrastructure gateway is optional and aimed at researchers, while emotional reaction changes with interaction frequency and content, showing happy, bored or sad states. The OLED shows seven expressions plus weather and time, and the speaker plays ten sound effects for boot, command confirm and emotion. The hardware list prices each module from Chinese shopping sites as of July 2026: an ESP12F module at 5 yuan, four SG90 servos at 12 yuan, an OLED at 6 yuan, an OV2640 camera at 15 yuan, and an HLK-V20 voice module at 9 yuan, with PCB and 3D printing essentially free using promotional coupons, totaling about 50 yuan. The prices are given as a real shopping list so a builder can replicate the dog without guessing component costs or sources. The prices are given as a real shopping list so a builder can replicate the dog without guessing component costs or hunting for sources.
Building and configuring
The README breaks replication into steps. First, flash the firmware using VSCode and PlatformIO: clone the repo, run pio run, upload, upload the web filesystem, then connect to the MindPaw WiFi and open the local address. Second, fabricate the PCB and 3D print the shell using promotional coupons from a Chinese board house. Third, solder and assemble, starting with power, then the core, then modules in order. Fourth, configure optional APIs for AI chat and weather, which the README says are not required for basic motion, expression and voice control. An optional AI infrastructure gateway can be started with Python and uvicorn, or with Docker Compose, and a mock provider lets the device link be verified without a cloud token. The voice module needs configuration on a PC with a tool, adding 22 voice commands before it is connected. The README also gives a wiring table and warns that GPIO15 must be pulled down with a 10K resistor or the ESP8266 will not boot, and that the camera lines sharing servos need a 1K resistor. Those two warnings are the kind of practical detail that saves a first time builder from a dead board. Those two wiring warnings are the kind of practical detail that saves a first time builder from a board that will not start or a camera that will not talk.
Licensing and structure
The repository is organized as a monorepo with a 3Dmodel folder, a SCH and PCB folder for hardware, a MindPaw_main folder for firmware, an ai-infra folder for the optional gateway, a Picture folder, and a Docs folder with guides for hardware, flashing, assembly, quick start and API setup. The license section states the software code is under the MIT License, the hardware design under CC BY-NC 4.0, the 3D models under CC BY-NC 4.0, and the documentation under CC BY 4.0. In plain terms, the code is free to use, while the hardware and 3D models may be used and modified personally but not commercially. The README answers common questions, such as whether a beginner can build it, whether the PCB or 3D print is required, and whether an ESP32 can substitute, with the answer that ESP32 is recommended for its extra pins and memory. It also notes the PCB can be hand soldered on a perf board and the shell can be made from cardboard or acrylic, lowering the barrier further. The project is hosted at github.com/ace-trump-tech/MindPaw and its most recent commit was on 2026-08-26. It also notes the PCB can be hand soldered on a perf board and the shell can be made from cardboard or acrylic, lowering the barrier further for newcomers.
Editorial conclusion
MindPaw is a desktop quadruped robot project built on an ESP8266 microcontroller, and it is hosted at github.com/ace-trump-tech/MindPaw with its most recent commit on 2026-08-26.
Community notes