Robotic Goalie

Robotics & Computer Vision · Autonomous Goal Defense
About

A real-time robotic goalie system that utilizes an iRobot Create 3 to autonomously defend a goal against incoming shots. Built on a Sense → Think → Act pipeline, the system streams live video from a robot-mounted smartphone to a laptop. The software processes the video feed to detect the incoming ball, uses a greedy centering algorithm to determine the necessary trajectory, and instantly issues precise movement commands to the robot via Bluetooth to block the shot.

Key Highlights
—Engineered a non-blocking software architecture using Python's asyncio and threading, ensuring the OpenCV video capture loop and the robot's Bluetooth event loop operated simultaneously without lagging.
—Built a robust data pipeline to merge, re-index, and manage custom datasets sourced from Roboflow, then trained and fine-tuned a YOLOv8 object detection model to reliably track high-speed targets.
—Bridged high-level computer vision logic with physical hardware execution, translating bounding-box coordinates into actionable ROS 2 commands via the iRobot Education SDK.
Skills
PythonYOLOv8OpenCVRoboflowiRobot SDKasyncioBluetooth BLE
Images
YOLOv8 object detection in action — the model identifies and tracks the incoming ball in real time, drawing a bounding box around the target.

YOLOv8 object detection in action — the model identifies and tracks the incoming ball in real time, drawing a bounding box around the target.

Overhead view of the robotic goalie setup — the iRobot Create 3 positioned in front of the goal, with the smartphone camera mounted to stream live video to the detection pipeline.

Overhead view of the robotic goalie setup — the iRobot Create 3 positioned in front of the goal, with the smartphone camera mounted to stream live video to the detection pipeline.