Whereas in the past ultrasonic sensors have been largely used only to estimate the proximity of objects and the location and identification of primitive targets in a robot workspace, the development of biomimetic sonar has opened up new possibilities for their application. Broadband sonar echoes have sufficient resolution so that characteristics on reflection, especially geometry and texture, can be distinguished with only a few measurements. In this paper, we describe how a model of texture can be used to distinguish between a number of different surfaces using only a single measurement of each, showing results on a number of surfaces that might be considered typical pathways for a mobile robot, both those with periodicity in pattern and those with statistically homogeneous features. In particular, we consider textures corresponding to hard smooth floors, carpets and asphalts, and surfaces with a repeating pattern made up of tiles. Each random rough surface is modeled using an extension of the Kirchhoff approximation method describing the scattering of the acoustic wave on the surface while the periodic surfaces are modeled assuming distinctive reflections from the tile borders. The continuous transmission frequency-modulated sonar signature corresponding to each class is derived and compared with the experimental measurement. A set of features is extracted that exploits the differences between the surface models, and a hierarchical classification scheme is proposed for recognition.

No abstract

Abstmct-The typical use of ultrasonic sensors has been limited to estimation of the location of targets in a robot workspace. C T F M sonars have also been used successfully in classifying primitive targets. In this paper the classification is extended to include textures typical of these found in pathways the robot m a y need to follow or identify. The pathway classes examined are considered t o be plane surfaces of various roughness corresponding to hard smooth poor, carpet, and asphalt. Each class is modeled using an extension of the Kirchhoff approximation method describing the scattering of the acoustic wave on rough surfaces. The C T F M sonar image corresponding to each class is derived and compared with the experimental one. Then a feature is extracted that exploits the differences between the three surface models. A neural network is trained for recognition with excellent results.