Sequential sensing with biosonar for natural landmark classification

Abstract

Echolocating bats can make nocturnal flights in acoustically cluttered environments with the use of echolocation. Their ability to evaluate targets in complete darkness provides mobile robots an opportunity to learn target detection, classification and identification with similar biomimetic platforms. In this work, natural landmark classification with a binaural system, a sequential sensing strategy and a frequency after reconstruction algorithm were developed and tested. The aim of the work is to overcome some inherent shortcomings of airborne sonar and take advantage of bats' perceived properties for mobile robots' navigation in natural environments. Experimental results suggest considerable improvements in classification accuracy can be achieved by the use of this sequential classification method.