AI-enabled robotic sorting for circular textile waste management: A scalable solution for India’s recycling sector

The global textile industry faces a critical inflexion point as circular economy mandates intensify and waste volumes soar beyond 100 million tonnes annually. Central to realising circularity is the efficiency and fidelity of textile waste sorting, a longstanding bottleneck dominated by manual, low-throughput, and error-prone methods. This paper investigates the deployment of an AI-enabled robotic sorting system integrating hyperspectral imaging (HSI) and deep learning algorithms within the context of India’s fragmented textile recycling ecosystem. We demonstrate that spectral imaging combined with convolutional neural networks (CNNs) achieves over 95% classification accuracy across heterogeneous, post-consumer Indian textile waste streams, including multi-fibre blends that typically confound manual sorters. Drawing from industrial benchmarks such as Sweden’s SipTex and U.S.-based Refiberd, we design a prototype that integrates conveyor automation, real-time classification, and robotic actuation. Comparative analysis reveals that the AI system achieves throughput rates exceeding 1,000 garments per hour, representing a 20× gain over manual processes while reducing misclassification rates by more than 60%. A techno-economic model suggests payback periods under four years when scaled to medium-sized facilities, with significant reductions in labour dependency and waste-to-landfill ratios. Our findings have strong implications for policy and industry: AI sorting systems not only align with India’s National Textile Policy and MITRA initiatives but also represent an enabling infrastructure for chemical recycling, extended producer responsibility, and traceable material flows. By bridging technological innovation with operational scalability, this study advances the industrial feasibility of circular textiles in the Global South.