Application of adaptive neural network to localization of objects using pressure array transducer

Pattern recognition and object localization, using various sensors such as vision and tactile sensors, are two important areas in the application of robotic systems. This paper demonstrates the feasibility of using some relatively inexpensive pressure sensors and a neural network to achieve object localization and pattern recognition. The sensors used are force sensing resistors (FSRs), more specifically, a 16×16 array of FSRs. Because of the nonlinearities associated with a FSR, three approaches for gathering output from the sensor array are used. The neural network used consists of two 2-layer counterpropagation networks (CPNs). In addition to recognizing pre-trained patterns, this paper also demonstrates that the conventional CPN configuration can be modified to learn new patterns even when its training period is completed. Both simulated and experimental results of this paper suggest that the neural network can provide an alternative approach for object localization using tactile arrays