Geo-referenced proximity detection of wildlife with WildScope: design and characterization

Abstract

Existing systems for wildlife monitoring focus either on acquiring the location of animals via GPS or detecting their proximity via wireless communication; the integration of the two, remarkably increasing the biological value of the data gathered, is hitherto unexplored. We offer this integration as our first contribution, embodied by our WildScope system whose key functionality is geo-referenced proximity detection of an animal to others or to landmarks. However, to be truly useful to biologists, the in-field monitoring system must be complemented by two key elements, largely neglected by the literature and constituting our other contributions: i) a model exposing the tradeoffs between accuracy and lifetime, enabling biologists to determine the configuration best suited to their needs, a task complicated by the rich set of on-board devices (GPS, low-power radio, GSM modem) whose activation depends strongly on the biological questions and target species at hand; ii) a validation in controlled experiments that, by eliciting the relationship between proximity detection, the distance at which it reliably occurs, and the location acquisition, provides the cornerstone for the biologists' analysis of wildlife behavior. We test WildScope in real-world experimental setups and deployments with different degrees of control, ascertaining the platform accuracy w.r.t. ground truth and comparing against a commercial proximity logger.