It is a common sight at public parks, schools and housing complexes: waste scattered on the grass just a few feet away from an empty dustbin. While most people simply complain about the lack of civic sense, 17-year-old Mahi Malhani decided to use artificial intelligence (AI) to solve the problem. Mahi, a Class 12 student at Amity International School in Mayur Vihar, Delhi, has built TRASHbot – a mobile, autonomous robot that rolls around public spaces, searches for litter and automatically separates garbage into the correct bins.
‘Brain’ behind TRASHbot
Unlike static, traditional recycling bins that rely on human, TRASHbot is designed and built with an AI powered autonomous trashbin built using Raspberry Pi 3b+, the AI brain and Arduino uno r3, as per the website.Raspberry Pi is a mini-computer that handles the heavy processing. Connected to a digital camera, it runs YOLOv3, a powerful real-time object detection AI library. When someone places trash near the robot, the camera takes a photo, identifies the item and classifies it.Once the Raspberry Pi identifies the trash, Arduino Uno signals the Arduino micro-controller, which commands two specialised turning motors to physically tilt the internal trapdoor, dropping the item into the correct compartment.“TRASHbot gives real time information and segregates waste automatically based on image classification by the YOLOv3 object detection library and classifies it into biodegradable and non-biodegradable. A mobile app helps in navigation of the bin, gamifying waste disposal methods. It aligns with UN Sustainable Development Goals (UNSDGs) 9 and 11, the website says.The robot operates both autonomously and via Bluetooth (manual control for events/demos), and it is made from lightweight foam board chassis, stable motor system and easily replaceable parts for durability.
Sorting trash with 90% accuracy
TRASHbot eliminates the confusing chore of manual recycling as it uses image classification, the machine instantly divides waste into two main categories: biodegradable (like food scraps and fruit peels) and non-biodegradable (like plastic wrappers and tin cans).According to the website, the robot has achieved 85% waste classification accuracy, 90% sorting mechanism success rate, 98% obstacle avoidance response rate and 92% real time classification of waste. On a single charge, the robot is claimed to offer patrolling a crowded area for two and a half hours. Target deployments are schools, offices, cafeterias and public spaces.