Drawing pig feeding patterns: sensor-driven algorithms for individual welfare monitoring

Pig welfare evaluation is currently limited to group-level assessment at one or a few moments in time. Animal welfare, however, is a dynamic process and differs between individuals housed in the same environment. Continuous monitoring of individual pig welfare can be achieved using Precision Livestock Farming (PLF) technology, which combines sensors and algorithms to detect deviations from basal (i.e. 'normal') pig behaviour; deviations that can be indicative welfare issues. Especially promising here are electronic feeding stations, which measure a wide range of feeding behaviours (e.g. intake, duration, frequency and rate) at different time scales, and for some of which links with pig welfare have already been established (e.g. reduced feeding frequency during lameness or reduced feed intake during heat stress). Nevertheless, the development of algorithms that can validly detect these deviations from basal feeding behaviour is currently hampered by a lack of understanding of the basal variation in temporal pig behaviour, making it difficult to isolate welfare-relevant from 'normal' variation. This project dives into this variation in feeding behaviour, both between animals and across time (day-to-day and as pigs grow), aiming to identify promising features of feeding behaviour from which welfare-relevant variation can be detected. These features will subsequently be used to develop algorithms that can automatically monitor individual pig welfare using electronic feeding station data.