A 16 Year Old Makes a Vision Sensor Bionic Hand Under $28

Bionic Hand Version 1: Vision-Controlled Robotics Under $28

I Am Dhruv ,At 16 years old, I developed Bionic Hand Version 1 to prove that advanced robotics doesn't require a high-end lab or expensive equipment. This project utilizes a standard laptop camera as a "vision sensor" to track human hand movements in real-time and translate them into mechanical action. It Took 3 months to make this after 5 failed iterations.

The Challenge of Resourcefulness:

A key highlight of this project is the construction method. To ensure this design remains accessible to students and makers without a 3D printer, I handcrafted the entire chassis from Sunboard (foamboard). This material provided the perfect balance of being lightweight for the motors and easy to modify during the prototyping phase.

How it Works:

Unlike traditional robotic hands that rely on bulky physical controllers, this system uses a "mirror" approach:

Vision Processing: Using Python, OpenCV, and MediaPipe, the software tracks 21 hand landmarks via a webcam.

Mechanical Sync: The software calculates the "Open/Closed" state of each finger and sends serialized data to an Arduino UNO.

Tendon-Driven Motion: High-strength fishing lines act as "synthetic tendons," pulling hand-carved trapezoidal finger segments into a natural grip

The Engineering Journey :

This project was the result of two months of intensive trial, error, and five failed versions. From dealing with mechanical friction to perfecting the serial communication between Python and hardware, every challenge was a lesson in persistence. By optimizing my materials, I successfully kept the total build cost under ₹2,500 ($28 USD).

I am entering this in the Sensors Contest to showcase how computer vision can be used as a powerful, non-contact sensor for assistive technology. My goal is to use the prize money to transition to 3D-printed joints in Version 2 to solve the mechanical friction limits of handmade foam.

The total build cost for Bionic Hand Version 1 was successfully kept under ₹2,157 (approximately $28 USD). This budget-friendly approach was achieved by utilizing affordable hobbyist components, such as MG90S metal-gear servos for durability and an Arduino UNO as the primary controller. By hand-carving the chassis from Sunboard (foamboard) instead of using expensive 3D printing services, the structural costs were minimized to just a few dollars. Power management was handled efficiently using a 2S Lithium-ion battery setup and an LM2596 Buck Converter, ensuring a stable 5V supply for the actuators without a high price tag.

Microcontroller: Arduino UNO — ₹259

Actuators: 5x MG90S Metal-Gear Servos — ₹600

Structure: Sunboard / Foamboard (25" x 20") — ₹250

Tendons: 20 Meters Nylon Fishing Line — ₹258

Power Supply: 2x 1200mAh 3.7V Lithium-ion Batteries with 2S Holder — ₹150

Voltage Regulation: LM2596 Buck Converter Module — ₹60

Charging: USB-C 2S Step-Up Boost Charger Module — ₹175 Connecting Hardware: Jumper Wires and 18-21 AWG Power Wires — ₹345

Mechanical Tensioners: 10x 3-inch Rubber Bands — ₹50

Finishing Materials: Electrical Tape and Sand Paper — ₹70

Total Estimated Cost: ₹2,157 ($28 USD)

The most important part of this bionic hand is the finger mechanism. I designed these to mimic human anatomy without using expensive mechanical hinges. Each finger is hand carved from Sunboard because it is lightweight and easy to shape.

The Trapezoidal Geometry I carved each finger segment into a specific trapezoidal shape. This design is very important because it allows the segments to sit flat when the hand is open but creates a natural stop when they are pulled. This mimics the way a human knuckle has a limit to how far it can bend.

Internal Tendon Routing To make sure the fingers move in a straight line I created an internal routing system. I used a small tool to hollow out two parallel channels inside the Sunboard segments. I then ran high strength nylon fishing line through these paths. This acts as the flexor tendon and keeps the movement stable without the finger twisting.

Passive Reset System To get the fingers to open back up I used rubber bands on the back of the hand. These act as elastic ligaments. When the servo motor pulls the string the finger curls. When the servo releases the tension the rubber bands automatically pull the finger back to the straight position.

The palm and arm chassis act as the foundation for the entire robotic system. Because I bypassed 3D printing I had to manually engineer a structure that could house five high torque servos while remaining thin enough to look realistic.

The Slim Palm Design I constructed the palm by layering sheets of Sunboard to create a rigid but lightweight frame. It is important to keep the palm profile as thin as possible. A thinner palm reduces the rotational inertia and lowers the amount of torque required from the motors. This allow the hand to respond instantly to the vision tracking software and creates a more seamless mirror effect between your movements and the robot.

Servo Staging and Alignment The arm chassis is built as an extension of the palm. I mounted five MG90S metal gear servos in a staggered formation along the forearm section. This specific layout is critical because it ensures that all five nylon tendons have a clear straight path from the servo pulleys to the finger joints. If the servos are not aligned correctly the strings can rub against each other which creates mechanical friction and can cause the servos to overheat or the strings to snap.

Integrating the Tendon Channels To connect the fingers to the body I routed the nylon fishing lines through the palm and down to the servos. I used the dual channel system to make sure each string pulls exactly from the center of each finger. By hand carving small guide paths into the Sunboard palm And Used IV Tubes(You Can Get Used IV Tubes From Any Hospital, I Managed To Get Them From A Hospital When I Had Jaundice )I ensured the tendons stay in their designated tracks even during rapid movements.

To translate the circular movement of the servos into the linear pull required for the fingers, I developed a custom pulley system. These pulleys are designed to maximize the "throw" of the servo, ensuring the fingers can move from fully open to a complete fist.

1. Cutting the Pulley Blanks Start by using a compass to draw circles with a 1.5cm radius on a sheet of Sunboard. You will need two circles for every finger, creating a total of ten discs. Cut these out carefully using a utility knife. To ensure they are perfectly balanced, I recommend stacking the discs and using medium-grit sandpaper to smooth the edges until they are uniform.

2. Creating the Tendon Groove To keep the nylon fishing line from slipping off the pulley, you must create a guide groove. Insert a toothpick or a thin piece of wire at the outer edge to create a small gap. This gap becomes the "track" where the tendon will sit as the servo rotates.

3. Mounting to the Servo Horns Once the pulleys are dry, you need to attach them to the plastic servo horns provided with your MG90S motors.

1. Use a soldering iron or a heated needle to create a small hole in the exact center of the Sunboard pulley.
2. Align the servo horn over this hole and use a high-strength glue to bond the plastic to the Sunboard.
3. For extra security, you can use the small screws that come with the servos to fix the horn firmly into the Sunboard material.

4. Routing the Strings Pass the nylon fishing line through the groove in the pulley. I recommend drilling a tiny hole from the outer groove to the center of the pulley and knotting the string there. This ensures that as the pulley turns, it "winds" the string onto itself, providing a consistent and powerful pull.

The electrical system is the nervous system of the bionic hand. Because this project uses five high-torque MG90S metal-gear servos, a standard Arduino power output is not enough. I designed a dedicated power rail system to ensure the motors receive constant voltage without crashing the microcontroller.

The Power Supply Chain I used two 18650 Lithium-ion batteries connected in series to create a 7.4V power source. To make the hand rechargeable and easy to use, I integrated a USB-C step-up boost charger module. This allows the entire hand to be charged using a standard phone charger.

Voltage Regulation with Buck Converter Since the Arduino and servos require a stable 5V to 6V, I used an LM2596 Buck Converter to step down the 7.4V battery voltage. This component is vital because it prevents the servos from drawing too much current directly from the Arduino, which would cause the board to reset or overheat. I tuned the converter to output exactly 5V for the best balance of speed and torque.

Wiring and Common Ground The wiring follows a parallel power architecture. The positive and negative outputs from the Buck Converter go to a power rail of Half Breadboard. All five servos connect their power and ground wires to this rail. It is extremely important to connect the ground of the battery system to the GND pin on the Arduino. This common ground ensures that the PWM signals from the Arduino have a clear reference point to control the servos correctly.

Control Signal Routing The signal wires from the five servos are connected to the digital PWM pins on the Arduino UNO. The Arduino acts as the bridge, receiving commands from the Python vision script via a USB cable and translating them into the precise movements for each finger.

The software for this project is divided into two parts that communicate in real-time. The first part is a Python script that act as the brain and the second part is an Arduino sketch that acts as the muscle.

Python and Computer Vision I used Python 3.9 along with the MediaPipe and OpenCV libraries to create the vision sensor. The script captures video from the laptop webcam and identifies 21 specific landmarks on the human hand. By calculating the distance between the tip of the finger and the base of the palm the code determines if a finger is open or closed. I optimized the code to run at a high frame rate so there is minimal lag between my movement and the robotic response.

Serial Communication Protocol To send data from the laptop to the Arduino I implemented serial communication at a baud rate of 115200. This high speed is necessary for smooth motion. The Python script sends a simple string of characters representing the state of each finger. For example sending a specific trigger word tells the Arduino to move a specific servo to a preset angle.

Arduino Logic and Servo Control The Arduino UNO runs a script that constantly listens for data on the serial port. When it receives a command it uses the Servo library to write a PWM signal to the corresponding MG90S motor. I carefully calibrated the servo angles in the code to ensure the motors do not pull the nylon strings too far which could snap the Sunboard chassis or strip the gears.

Note : If You Wish To Make The Whole Structure Using A 3D Printer, You Should Definately Do So. Because I Don't Have One So I Used The Resources I Have At This Time. If You Make The Structure With The 3D Printer You Can Achieve More Smooth Movement. You Can Use Same Code, Electronics, And Method In That.....

This project isn't perfect, but it proves that high-end tech can be built on a student’s budget. I’ve learned so much during this build, and I promise to keep improving the design, making it even more precise and durable in the future. Stay tuned for the next version!