Tinker is an open-source bipedal robot designed to provide robotic enthusiasts and developers with a hands-on platform. This project includes a detailed list of required hardware materials, software resources, and comprehensive installation steps to help users easily build and program their own robot.
Through Tinker, users will not only gain insights into the fundamental principles of robot movement and control, but also have the opportunity to personalize and enhance their projects. We aim to inspire more people to explore and engage with robotics technology through this initiative.
🚶 Reinforcement Learning-Based Control
🧾 Step-by-Step Assembly Instructions
💡 Customization and Expansion
- Reinforcement Learning Algorithm Adaptation
The robot is specifically designed for reinforcement learning algorithms, enabling autonomous learning and continuous optimization of action strategies in complex and dynamic environments. Through deep integration with reinforcement learning, it demonstrates outstanding performance in motion control, task execution, and action planning. For example, when facing unknown terrains or complex tasks, the robot can quickly adapt and make reasonable decisions to efficiently complete tasks.
- Direct Deployment and Testing
This system features a convenient direct deployment and testing function, making it widely applicable to various task scenarios. By simply connecting the device, users can quickly initiate robot testing, significantly reducing the R&D cycle and improving work efficiency. Compared to traditional deployment methods, it offers significant advantages in autonomy, flexibility, cost control, adaptability, and safety.
- Complete Solution from Simulation to Real-World Deployment
The solution includes IsaacGym-based RL simulation, sim-to-sim transfer, real-world deployment, and secondary development tutorials.
- Open-Source Training Software
The system is equipped with open-source training software that provides a wide range of functional modules, including model training settings, data visualization, and algorithm debugging interfaces. Users can freely adjust training parameters, monitor data changes in real-time during training, and optimize algorithms while troubleshooting issues. The open-source nature offers great flexibility and scalability, allowing users to customize training plans according to their needs. Additionally, the open-source community provides valuable resources and technical exchange opportunities to accelerate development.
- Integrated Solution: Physical Robot + Reinforcement Learning
This solution combines real-world robotics with reinforcement learning to provide a comprehensive development and deployment framework.
-
Motion Performance
- Maximum Walking Speed: The robot can achieve a stable walking speed of up to 1m/s and a normal walking speed of 0.5m/s, demonstrating excellent mobility. It can move quickly through various environments, meeting the speed requirements for multiple application scenarios.
- Load Capacity: The robot has a load capacity of 1kg, allowing it to carry tools or objects and perform tasks such as item transportation and equipment operation.
-
Battery Life
- Battery: The robot is powered by a 21.6V 5Ah 108Wh battery with a charging voltage of 25.2V and comes with a dedicated charger. The battery life is yet to be measured, aiming to provide stable power for long-duration operations.
- Energy-Saving Design: The system incorporates energy-saving optimizations at both hardware and software levels to reduce power consumption, extend battery life, and improve overall efficiency.
-
Control System
- Controller: Equipped with high-performance controllers such as NVIDIA Jetson Nano, Odroid, and STM32, enabling fast data processing for real-time response and precise control.
- Sensors: The system integrates advanced sensors, including a gyroscope, accelerometer, and electronic compass, to continuously monitor the robot’s posture, motion state, and surrounding environment. This provides essential data for accurate control and intelligent decision-making. Additionally, it supports Wi-Fi and 2.4G wireless Bluetooth, ensuring convenient data transmission and remote control capabilities.
This section will provide a step-by-step guide on how to operate and interact with the Tinker robot once assembled.
A detailed Bill of Materials (BOM) required to build the Tinker robot will be listed here. It includes hardware components, sensors, and other necessary parts.
Comprehensive instructions on how to assemble Tinker from scratch, complete with images, will be made available in this section.
Explore how to extend and enhance Tinker’s functionalities through custom development. This section will cover software development.
YouTube:
bilibili:
Find more process videos here.
AGPL-3.0
✨Feel free to customize this further based on specific content for each section!✨
For more open-source information, visit OpenLoong.