Distributed sensor swarms for monitoring bird behavior: an integrated system using wildlife acoustics recorders

Abstract

Birds are frequently vocal. Monitoring their songs and calls has provided much useful information about their ecology and behavior. Recording bird songs and locations can be resource intensive, frequently requiring two or more observers, causing considerable disturbance and possible only for short times. Passive arrays of acoustic sensors offer the possibility of greatly increasing our ability to monitor bird activity. The desirability of such arrays is obvious: they are less intrusive, can monitor continuously over long periods, permit collaboration which enables better localization, provide fault tolerance, and facilitate sharing to optimize scarce resources. That the interest has been increasing so much lately is in large part, because we are now at the point, where the promise of such arrays is realizable with current or reasonably anticipated technologies. Controlling collaborative arrays can be difficult. Engineered solutions are sometimes available, but they are often brittle and appropriate only for ideal environments. We would like our systems to be: robust, in that they can handle changing environments or agents and are untroubled by occasionally wrong or noisy messages; adaptive in that they can learn to deal with unanticipated source or events, form new concepts, and communicate in languages that are specialized for particular agents; and finally, they should be self-configuring, to deal with changing situations and goals. After reviewing research employing sensor arrays by others to monitor bird behavior, we will describe research in our laboratory. Our focus will be on the construction of such systems, especially how such arrays can extract information from the environment and communicate to arrive at a collective understanding of their region and events occurring there.