An Autonomous System for the Self-supervision of Animal Fattening in the context of Precision Livestock Farming

Abstract

Beef production needs certain levels of autonomy to ensure that animal fattening processes achieve certain sustainability objectives (e.g., financial and environmental). For example, it is required oversight in the animal fattening process, so that stakeholders can make better decisions about what is happening in the fattening process. For monitoring the animal fattening process, this paper proposes an autonomous system. In this paper, this autonomous system is designed and implemented using the methodology for the development of data Mining applications called MIDANO, and is tested in a cattle farm simulator that has been developed to reproduce the events of the animal fattening production process. This autonomous system for the self-supervision of the animal fattening process is composed of two data analysis tasks, one to detect anomalies in the fattening of cattle, and another to diagnose this anomaly. The results with real data demonstrate the ability of the proposed supervision system to detect and diagnose anomalies in various conditions (normal, animal health problems, and forage problems in the paddock), and the possible causes of abnormal values in the weight variable. The anomaly detection models have a MAE of the order of 5.5 kg, and the diagnostic model has 95% of Accuracy and 1 of AUC. The results of the experiments are encouraging, as they show that the autonomous system is capable of detecting anomalies and diagnosing them in different operating scenarios. Our system allows giving self-supervision characteristics to a production process.