318 — A Hidden Markov Model to Estimate the Time Dairy Cows Spend in Feeder Based on Indoor Positioning Data

Pastell & Frondelius (10.1101/250092)

Read on 04 July 2018
#markov-model  #dairy  #cow  #farm  #GIS  #feeding 

The times at which dairy cows choose to feed can often be an indicator of health or other metabolic changes; a hungry or undernourished cow is a sad cow, which is a statement that also applies to other animals including me.

When a domesticated inside-cow is positioned at a feeding trough, it’s almost certainly eating. So determining cow position and orientation is virtually adequate to determine how much a cow is eating. Using an Ubisense indoor positioning system, the authors looked at how much different dairy cow individuals ate from a feed bunk during a period of time.

Of the 50 dairy cows measured in this experimental freestall barn (a barn with stalls available for cows to voluntarily occupy), a Markov model was adequate to predict when a cow would choose to feed, with 97.6% accuracy.

Though it’s not feasible to install this positioning system technology in every barn, it’s an interesting study that enables us to reduce the complexity of the cow-feeding and cow-nutrition problem to a much simpler problem (by proxy of the Markov model).