AI on animals: AI-assisted animal-borne logger never misses the moments that biologists want

Abstract

Animal-borne data loggers, i.e., biologgers, allow researchers to record a variety of sensor data from animals in their natural environments (Hussey et al. 2015; Kays et al. 2015). This data allows biologists to observe many aspects of the animals lives, including their behavior, physiology, social interactions, and external environment. However, the need to limit the size of these devices to a small fraction of the animals size imposes strict limits on the devices hardware and battery capacities (Kays et al. 2015). Here we show how AI can be leveraged on board these devices to intelligently control their activation of costly sensors, e.g., video cameras, allowing them to make the most of their limited resources during long deployment periods. Our method goes beyond previous works that have proposed controlling such costly sensors using simple threshold-based triggers, e.g., depth-based (Watanuki et al. 2007; Volpov et al. 2015) and acceleration-based (Nishiumi et al. 2018; Brown et al. 2012) triggers. Using AI-assisted biologgers, biologists can focus their data collection on specific complex target behaviors such as foraging activities, allowing them to automatically record video that captures only the moments they want to see. By doing so, the biologger can reserve its battery power for recording only those target activities. We anticipate our work will provide motivation for more widespread adoption of AI techniques on biologgers, both for intelligent sensor control and intelligent onboard data processing. Such techniques can not only be used to control what is collected by such devices, but also what is transmitted off the devices, such as is done by satellite relay tags (Cox et al. 2018).