High-resolution monitoring of social interactions in dairy cows using a prototype Ultra-Wideband system

Social interactions significantly influence the welfare, health, and productivity of dairy cows. In intensive housing, sociality encompasses a spectrum of behaviours, from proximity and physical contact to nonphysical cues such as avoidance. The ability to detect these types of behaviour is crucial for enabling automated systems for welfare evaluation, productivity optimisation, and early-stage disease detection [1]. Indoor tracking of individual animals poses unique challenges due to line-of-sight obstructions for sensor-based systems and visual constraints in camera monitoring. After a preliminary evaluation under controlled conditions [2], a prototype Ultra-Wideband (UWB) system was selected for a field application on a commercial farm. This study aimed to validate the prototype system for indoor positioning of animals, to study inter-individual proximity and spatial utilisation patterns in permanently housed livestock. The system was composed of 12 prototype sensing units fitted on collars, which included a UWB module operating as a mobile node, a microcontroller, a real-time clock and an SD card, plus 8 fixed nodes (anchors). Absolute position data were collected for 6 consecutive days, at a sampling rate of 1Hz. Simultaneous video recordings were obtained using 4 time-lapse cameras covering the entire experimental area, recording 1 frame every 3 seconds. The twelve Holstein cows were housed in a pen that included permanent straw bedding covering the entire area, a drinking trough, and an external covered zone (total surface area 153 m2). Data were processed in a GIS environment. Video recordings were utilised to establish empirical inter-individual distance and temporal thresholds for defining contact events. These thresholds were used to filter UWB data, yielding clusters of dyadic interactions (Fig. 1). Furthermore, the frequency of positions in the feeding, resting, and drinking areas was obtained for each cow. Spatial distribution was quantified through Kernel Density Estimation and spatial autocorrelation analysis to evaluate the occupancy of feeding, resting, and drinking functional areas. The detection of proximity events using the UWB system showed strong alignment with camera-based monitoring. The integration of high-resolution positioning with spatial statistics allowed for a quantitative assessment of how cows distribute within the barn. The experiment demonstrated the potential for UWB technology to provide continuous, automated welfare evaluation by tracking social dynamics and resource utilisation in real time.