Full range of motion robotic hand with pressure, position and force sensors for $300 bill of materials.
Working on release candidate mk14. Using own motor drivers for fine force control, enhanced integrated sensor suite, completely refactored control code, easy USB connection, and improved manufacturing and repairability.
Patented design with integrated sensor suite! GB2626014
New design routes all cables and tendons through the center of the wrist, hiding them from view and increasing the range of movement.
The Hextech mechahand is designed to be a low cost robot hand with a similar range of motion as a human hand. It's targeted as an accessible standard for reinforcement learning research. It has 20 degrees of freedom with position and force feedback on each joint. The position encoders are placed in-situ allowing for accurate sensing of the hand state regardless of the stretch on the tendons. Additionally there are 6 pressure sensors in the finger tips and palm for finer feedback. All electronics plug into a compact PCB with current sensing for each actuated joint. The hand can be controlled remotely by wifi using a force-limited PID loop. In the video, it's mimicking my hand position in order to grasp an object. For autonomous control, a custom force aware loop can be implemented.
The hand is built using off the shelf components and a 3D printed body at an amortized cost of 300 USD per hand, requiring 30 hours of assembly time. Included in the cost is the custom PCB designed for automated assembly. All design files for the mechanism, the control circuit, and driver code is open source. Please open an issue wherever build instructions need to be clarified.
The project contains code for a 24 channel driver, the 3D printable model of the hand, and code to control the hand with a leap sensors. Details in the readme file of each folder.
- 24driver: Code and circuit design for a 24 channel PID controller with wireless interface and 12 pressure sensor inputs.
- hand_model: 3D model and instructions for the mecha-hand versions.
- leap_control: Code to control the hand via the leap motion sensor.