Enhancing cat care: Unveiling the technology of intelligent litter box monitoring

Abstract

While interest in feline elimination behavior in litter boxes tends to focus on factors related to the pain-points of pet cat care, there have been recent efforts to develop “smart” devices to track litter box activity as a means to observe and assess cat health. The intelligence of these devices is limited, however, given the complexity of measuring cat behavior within the constraints of the litter box environment. Here, we describe the successful development of an intelligent device that has proved a means to systematically document feline elimination patterns across a large, representative sample. The device is equipped with load cell sensors within a platform that is placed unobtrusively under a cat’s existing litter box. The rigorously developed AI models relied on a supervised learning methodology rooted in a feature generation module, all of which was made possible by a robust truth dataset of hundreds of thousands of carefully labeled litter box events captured on camera. The AI engine of this new tool can confidently distinguish Cat from Human events as well as identify the type of Cat event that is occurring (i.e., urination, defecation, non-elimination), unique cats, and duration metrics, among other event features. The performance of all models meets or exceeds the 80 % confidence threshold indicating that this device is a reliable tool that can be leveraged for future research into additional aspects of feline elimination.