NAS Media Limited Robot
OpenDuckMini
US$ 798.00
A low-cost, fully open-source bipedal biomimetic robot platform designed specifically for embodied intelligence and reinforcement learning research, aiming to enable university laboratories and individual developers to easily get started and freely reprod
Quantity
Product details
Product Features:
| Bipedal Dynamic Walking | Through the 16-degree-of-freedom servo system to coordinate joint movements, combined with gyroscopes and PID control, it achieves bipedal dynamic balance and humanoid gait walking. |
| Reinforcement Learning Behavior Training | Train gait and control strategies using algorithms such as PPO/SAC in simulation environments like Isaac Sim or MuJoCo, and transfer them to physical robots. |
| Complex Action Execution | Supports bionic movements such as squatting, standing up, and nodding, and can perform motion programming and trajectory planning (inverse kinematics solution to be improved). |
| Fall self-recovery | Detect the fall state through the gyroscope and trigger a pre-set sequence of recovery actions to stand up again (to be improved) |
| Visual Recognition | Based on the Raspberry Pi camera, implement face tracking, YOLO object detection, and ORB-SLAM3 simple scene mapping (to be improved). |
| Voice Interaction | Supports Vosk/Whisper voice command recognition and TTS voice synthesis, and can interface with large models to implement a dialogue system (to be improved). |
| Haptic Feedback | The collision is detected by the micro switch on the fuselage, triggering obstacle avoidance or stop actions. |
| Emotional Expression | Using LED eye masks to display emotional states such as joy and surprise enhances the affinity of human-machine interaction. |
| Autonomous Navigation | Combining visual SLAM with A*/DWA path planning algorithms, it explores and reaches the target point in an unknown environment (to be improved). |
| MultiModal Machine Learning Interaction Decision Making | Integrate speech and visual inputs to execute complex instructions such as "walk towards the person wearing red clothes and say hello" (to be improved). |
| Cloud Collaboration | Connect to Azure IoT or AWS RoboMaker to achieve remote monitoring, data postback, and cluster control (to be improved). |
| Hardware Expansion | It can be equipped with a lidar to improve mapping accuracy, or connected to a robotic arm to perform extended tasks such as grasping (to be improved). |
| Educational Research Scenario | Supports ROS 2 teaching, decision verification of embodied intelligence large models, and comparison of the physical transfer effects of different reinforcement learning algorithms. |
Standard Edition Product Specifications:
| Main Controller | Raspberry Pi Zero 2 W (quad-core 64-bit ARM Cortex-A53 @ 1GHz, 512MB RAM) + 16GB SD card |
| motion unit | 14 * Feite Servo ST3215-C018 (12V, 18kg·cm Torque) + 2 * SG90 Auxiliary Servo |
| Sensor System | BNO055 9-axis attitude sensor (gyroscope + accelerometer + magnetometer) Raspberry Pi Camera V2 (8 megapixels) |
| Power Management | Battery: 1500mAh 3S LiPo battery (11.1V) Step-down module: UBEC 12V→5V (power supply for Raspberry Pi) Driver board 12V: Supports 14-channel serial servo motors |
| Audio System | 40mm Pot Bottom Horn + MAX98357 I2S Digital Audio Amplifier |
| Communication Interface | Wi-Fi / Bluetooth (built-in Raspberry Pi), Type-C (data) |
| expression system | LED Eye Mask |
| Controller | Handle |
| structural component | 3D Printed Parts (PLA/PETG Material) - Legs * 2 / Torso * 1 / Head * 1 / Joint Connectors * 16 |
| Dimensions (standing) | Approximately 25cm (height) × 18cm (width) × 12cm (depth) |
| Weight | Approximately 1.2kg (including battery) |
| Joint Degrees of Freedom | 16 Degrees of Freedom (Legs × 6 / Side Swing × 4 / Head × 2) |
| Maximum movement speed | Walking speed is approximately 0.15m/s |
| Continuous working duration | Approximately 45 minutes (dynamic equilibrium mode) |
| Operating System | Raspberry Pi OS（Linux） + Python + ROS2（To be improved） |
| Simulation Platform | NVIDIA Isaac Sim / MuJoCo / PyBullet |
| Core Algorithm | Reinforcement Learning (PPO/SAC), PID Dynamic Balance Control, ORB-SLAM3 Visual Localization (Optional) |
| Pre-trained simulation model | Open source provides |
| AI capabilities | Voice Interaction: Vosk Speech Recognition + GPT Conversation Engine (to be improved) Vision: YOLO Object Detection/Face Tracking (to be improved) Emotional Feedback (LED Eye Mask Expression) |
| Development Interface | Python API / ROS 2 Node (to be improved) / Web Remote Control Interface (to be improved) |
| Hardware Expansion | Reserved I2C/SPI/UART interfaces (x6), compatible with LiDAR (RPLidar A1), and supports robotic arm add-on modules |
| Operating Voltage | Main Controller: 5V DC / Servo: 12V DC |
| Peak Power Consumption | Approximately 65W (all servos at full load) |
| Charging Specification | DC 12.6V/1A Balanced Charger |
| Low Voltage Protection | Automatic alarm when battery voltage < 10.5V |
| Pre-trained simulation model | Open source provides |
| Getting Started Tutorials and Videos | has |
| Official Documentation (English) | has |
| Official GITHUB Source Code Repository | has |
| Official CAD Model Library | has |
| Video Collection | has |
| Technical Support Duration | 1 week |
Simulation Effect Video:
© COPYRIGHT 2026 WEBER. ALL RIGHTS RESERVED