Robotics + Computer Vision + AI
An autonomous air hockey opponent powered by real-time computer vision, predictive AI, and a custom-built 2-axis robotic mechanism. Human vs. Machine, one puck at a time.
01 Overview
This project transforms a standard consumer air hockey table into a fully autonomous AI opponent. An overhead camera tracks the puck in real time, a prediction algorithm calculates the trajectory, and a custom-built 2-axis linear actuator system drives the AI's mallet to intercept, block, and strike back. The result: a machine that plays air hockey against a human, reacting in milliseconds.
02 Media
03 Technology
Overhead camera captures the playing field. Real-time image processing isolates the puck and tracks its position frame by frame with sub-pixel accuracy.
Calculates puck velocity and trajectory from tracked positions. Predicts bounce angles off walls and determines optimal intercept point for the AI mallet.
Custom 2-axis gantry built from linear rails and stepper motors. Controlled via microcontroller, translating predicted intercept coordinates into precise physical movement.
Beyond pure defense: the AI evaluates when to block, when to counter-attack, and how to position for advantage. Adaptive difficulty keeps the game competitive and fun.
04 How It Works
Overhead camera feeds live video of the playing surface to the processing unit at high frame rates.
Computer vision isolates the puck position from the background using color filtering and contour detection.
The algorithm calculates velocity, direction, and wall bounces to determine where the puck will arrive.
The 2-axis actuator moves the mallet to the intercept point, executing a block or counter-attack.
