Edge Computing for Cattle Behavior Analysis

peer reviewedSmartphones, particularly iPhone, can be relevantinstruments for researchers because they are widely used aroundthe world in multiple domains of applications such as animalbehavior. iPhone are readily available on the market, containmany sensors and require no hardware development. They areequipped with high performance inertial measurement units(IMU) and absolute positioning systems analyzing user’s move-ments, but they can easily be diverted to analyze likewise thebehaviors of domestic animals such as cattle. Using smartphonesto study animal behavior requires the improvement of theautonomy to allow the acquisition of many variables at a highfrequency over long periods of time on a large number ofindividuals for their further processing through various modelsand decision-making tools. Indeed, storing, treating data at theiPhone level with an optimal consumption of energy to maximizebattery life was achieved by using edge computing on the iPhone.This processing reduced the size of the raw data by 42% onaverage by eliminating redundancies. The decrease in samplingfrequency, the selection of the most important variables andpostponing calculations to the cloud allowed also an increasein battery life by reducing of amount of data to transmit. Inall these use cases, the lambda architectures were used to ingeststreaming time series data from the Internet of Things. Cattle,farm animals’ behavior consumes relevant data from InertialMeasurement Unit (IMU) transmitted or locally stored on thedevice. Data are discharged offline and then ingested by batchprocessing of the Lambda Architecture