Neuromorphic encoding strategies for a noisy magnetic sense.

Abstract

While accurate engineered solutions to determine global position require vast networks of well-maintained transponder stations, many animals can solve this problem using only Earth's magnetic field. Moreover, animals are capable of this feat despite evidence suggesting that the magnetic sense operates at an extremely low signal-to-noise ratio. As such, this sense may provide valuable insights for the engineer. Here, we explore neuromorphic encoding strategies that may underlie this ability in animals and test their ability to accurately encode noisy magnetic information. We describe sparse encoding strategies that may function in this sense, with systems composed of as few as eight receptors and tens of neurons. We also find that neural architecture based on the arthropod central complex (implicated in other orientation tasks) is particularly robust to encoding noisy magnetic field information.