Neural Processing of Auditory-tactile Sensor Data to Perform Reactive Behavior of Walking Machines

Spiders can sense sounds in a frequency range between approximately 40 and 600 Hz by the use of hairs; they can detect e.g. the puff of wind of buzzing flies. On the contrary, scorpions use hairs as tactile sensors for obstacle avoidance. To integrate the advantages of both types of sensoric hairs, this article presents an artificial auditory-tactile sensor system, which combines the principles of the auditory hairs of spiders and the tactile hairs of scorpions, and investigates some neural techniques for processing these sensor signals. The different types of signals are discerned by recurrent neural networks in such a way that their output can generate different reactive behavior, like obstacle avoidance and tropism, of a walking machine. An evolutionary algorithm is applied to find an appropriate solution to this problem.