Modeling Wireless Sensor Networks for Monitoring in Biological Processes

Management and control of a herd of animals relies on monitoring the herd, which is very complicated due to the geographical distribution of the animals in outdoor environments. Animal behavior monitoring systems in outdoor environments should be able to remotely register relevant behavioral parameters, as the use of wired sensors is impractical.

In this thesis, a ZigBee based wireless sensor network was employed and only a part of the herd was monitored, as monitoring each individual animal in a large herd under practical conditions is inefficient. Investigations to show that the monitored animals can indeed represent the whole herd were carried out. The tagged animals in the herd were equipped with wireless nodes around the neck capable of measuring two behavioral parameters: the pitch angle of the neck (using accelerometer) and the velocity of the movement of the animal (using received signal strength). Fusing the two measured behavioral data resulted in an improvement of the classification results regarding the animal behavior mode (activity/inactivity), compared to the results achieved by only monitoring one of the behavioral parameters. Applying a multiple-model adaptive estimation (MMAE) approach to the data resulted in the highest classification success rate, due to the use of precise forth-order mathematical models which relate the feed offer to the pitch angle of the neck.

This thesis shows that wireless sensor networks can be successfully employed to monitor the behavior of a herd of dairy cows in outdoor environments.