Design and Development of an AI-Powered Autonomous Waste Sorting Robot Using Faster R-CNN and Robotic Manipulation

The rapid growth of municipal solid waste (MSW) has also led to the need for automated intelligent waste management systems that enable high purity of recycling and reduce human exposure to hazardous waste. Manual sorting is inefficient and uneconomical and poses health hazards. This paper, on the design, fabrication, and control of an autonomous mobile robot integrated with a 4-DOF robotic arm and a vision system powered by AI for small waste collection and classification, is proposed. Waste can be simply separated into three general categories: plastic, paper, and metal. Instead of speeding up the traditional single-stage detector by sacrificing the localization precision, a two-stage Faster R-CNN system with a ResNet-101 backbone plus FPN is deployed. This is combined with OpenCV spatial mapping and RGB-D depth perception to guide the robotic manipulator. The mechanical design, kinematic modelling, electronic integration, and software architecture are described. The results of our experiments yield a classification accuracy of 94.1% and a physical sorting success rate of 90%, confirming the real-world viability of the system for micro-waste management.