💡 Inspiration
Robotics has a gatekeeping problem: Hardware is expensive. To build a high-fidelity teleoperation system (controlling a robot remotely), you typically need haptic gloves, exoskeletons, or complex master-slave controllers that cost thousands of dollars. This creates a massive barrier to entry for innovators in hazardous fields like nuclear cleanup, telesurgery, and space robotics.
I asked a simple question: Can we replace thousands of dollars of sensors with a single webcam and better math?
Mimicry.ai is the answer. It is a "Zero-Hardware" interface that democratizes robotic control, turning a standard laptop into a precision kinematic controller.
⚙️ How I built it
I built the system in Python using a modular architecture that separates "Perception" from "Control."
- Perception Layer: I used MediaPipe to extract high-frequency skeletal landmarks from the user's hand in real-time.
- Kinematic Core: Instead of simply mapping pixels to servos (which is dangerous and inaccurate), I wrote a custom Inverse Kinematics (IK) solver from scratch.
- Simulation Layer: I built a virtual 2-Link Planar Arm using OpenCV to visualize the mechanical response instantly.
🧮 The Math (Inverse Kinematics)
The core innovation is the physics engine. To ensure the robot moves naturally, I solve the geometric configuration of the arm using the Law of Cosines.
Given a target coordinate $$(x, y)$$, we calculate the elbow angle $$\theta_2$$:
$$\theta_2 = \arccos\left(\frac{x^2 + y^2 - L_1^2 - L_2^2}{2 L_1 L_2}\right)$$
Once the elbow angle is known, we resolve the shoulder angle $$\theta_1$$ by correcting the geometric arctangent with the offset created by the second link:
$$\theta_1 = \arctan\left(\frac{y}{x}\right) - \arctan\left(\frac{L_2 \sin(\theta_2)}{L_1 + L_2 \cos(\theta_2)}\right)$$
This ensures that the software outputs valid angles that can be sent directly to physical stepper motors or servos (via ESP32) without
Log in or sign up for Devpost to join the conversation.