Filling the Gaps: Optimising Interpolation Methods for Missing Positioning Data in Dairy Cattle

Background

Real-time positioning systems are becoming a cornerstone of precision livestock farming, with applications ranging from behavioural monitoring to early disease detection. Several commercial systems are currently available, including DeLaval’s BA300, Lely’s positioning solutions, and Växa’s system. Despite their potential, these technologies face a critical limitation: they do not record 100% of the animals’ positions. In fact, our research group has shown that, on average, approximately seven hours of data per animal per day are missing, often in short gaps of around five seconds.

These gaps interfere with the accurate measurement of movement patterns and behavioural indicators, which are essential for health monitoring and farm management decisions. Interpolation methods, which estimate missing positions from surrounding data points, therefore play a vital role in ensuring reliable outputs. Previous work from our group has tested individual interpolation algorithms (see Ren et al. 2022), but the potential of combining methods to maximize accuracy has not yet been fully explored.

Aim

The aim of this project is to identify and evaluate the most effective combination of interpolation methods to reconstruct missing positioning data in dairy cattle. By testing and optimising different algorithms, the study will improve the accuracy of behavioural metrics derived from positioning systems and support the development of reliable monitoring tools in precision livestock farming.

Project description

This project will focus on:

• Cleaning and processing positioning data from sensor-equipped dairy cows
• Testing different interpolation algorithms to reconstruct missing data
• Developing and evaluating combinations of methods to improve accuracy
• Assessing the impact of improved interpolation on behavioural and movement metrics

Learning Outcomes

• Practical experience in R programming for data cleaning, interpolation, and validation
• Skills in evaluating algorithm performance and statistical accuracy
• Understanding of how positioning technologies support animal welfare and health monitoring
• Insights into the design of next-generation tools for precision livestock farming

Specifications

• Prior experience with R or another programming language is an advantage, but not required. As the analysis will be performed using a custom-developed R package, full supervision will be provided.
• Suitable for students in animal science, veterinary medicine, data science, or related fields with an interest in smart farming technologies and animal behaviour.