Vision-Guided Sorting Robotic Arm
This advanced curriculum capstone connects a camera observation to a physical action. A model arm sorts coloured blocks into separate bins. The challenge is not simply to recognise a colour: the student must coordinate perception, movement and safe stopping.
Suitable grade and prerequisites
The existing Grades 9-10 pathway, approximately ages 14-16, covers Python, NumPy, OpenCV and robotic-arm servos. Students benefit from experience with variables, conditions, functions and basic circuits. The demo can help identify whether the student needs a foundation stage before beginning this capstone.
Components and the system boundary
The curriculum lists Raspberry Pi 4 or ESP32-CAM, camera hardware, a four-degree-of-freedom servo arm and a chassis kit. The described sorting example uses red and blue cubes. The exact camera/controller arrangement is chosen by the mentor; not every board runs the same software stack. The arm operates in a supervised model workspace, not an industrial environment.
Vision, code and motion
Start with a still image and test colour thresholds under different lighting. Python and OpenCV can isolate candidate regions before a movement is requested. This rule-based vision stage should not be described as a trained neural network. A later classifier can be compared with it when the learner is ready.
Separate camera processing from servo commands. The program needs a clear sequence for detecting, positioning, gripping, moving and releasing, plus a stop response for uncertain input. Test the arm without a payload first, then introduce blocks. Mechanical limits and power supply behaviour can explain failures that look like software bugs.
Evidence of learning
A useful demonstration includes an unfamiliar lighting condition or a partly hidden block, not only the easiest case. Students explain when the system should refuse to move and how they checked the result. The outcome is a documented engineering investigation, not a promise of perfect recognition. Photos and short videos are pending genuine builds after opening.