ZapSortBot: Autonomous robotic sorting of used batteries utilizing deep learning

The increasing volume of spent batteries from consumer electronics and electric vehicles poses significant environmental and safety challenges for recycling systems. Efficient battery sorting by chemistry and size is critical for ensuring safe downstream processing and maximizing material recovery. This paper presents a low-cost, modular system that classifies and sorts batteries using a multi-modal approach combining AI-powered vision and weight-based inference. ZapSortBot employs a YOLO-v8 object detection model for real-time localization, with downstream logic determining battery chemistry type based on visual and physical cues. A 6-DoF robot (Niryo Ned 2) and a conveyor belt have been used to perform the sorting. The system achieved a classification accuracy of 98.3% across four common chemistries (alkaline, nickel–metal hydride, lithium, and zinc–carbon), with robust performance under varying lighting and battery conditions. Compared to high-cost industrial systems relying on spectroscopic techniques or X-ray imaging, ZapSortBot offers an accessible alternative for decentralized recycling environments. The system’s modularity, retainability, and special-case handling further support deployment in resource-constrained settings, addressing a critical gap in scalable, intelligent battery sorting technology.