Autonomous Recycling Sorter

Around 25% of recycling ends up contaminated and disposed of in a landfill. This project aimed to mitigate recycling contamination by autonomously sorting trash from recycling, so that no human can accidentally contaminate the recycling. This was a course project for my Intro. to Engineering Design course at RPI with a multidisciplinary team of engineering students. 

​

Our team developed a device to classify items as trash or recycling through a camera module, leveraging the power of machine learning with the potential for scaling with the vast amounts of image data on the internet. This device also consisted of a conveyor belt and two rotating flaps to sort items into their respective bins. 

​

My role consisted of embedded programming, electrical wiring, and design of the identification subsystem. I wrote stepper motor control code, trained a CNN classifier in TensorFlow and ported it to run on a Raspberry Pi. 

​

We faced challenges with mechanical alignment between subsystems, especially those with shafts driven by motors. Additionally, developing a classifier whose prediction is stable to a changing background (what happens when the lighting changes, or the camera sees below the rotating flap, see video) was a significant technical challenge that we overcame. 

​