Towards integrating intelligence in electronic skin. (March 2016)

• Record Type: Journal Article
• Authors: Seminara, Lucia; Pinna, Luigi; Ibrahim, Ali; Noli, Luca; Caviglia, Stefano; Gastaldo, Paolo; Valle, Maurizio
• Is Part Of: Mechatronics. Volume 34(2016)
• Page Start: 84
• Page End: 94
• Publication Date: 2016-03
• Subjects: Materials integrated sensing -- E-skin -- Piezoelectric polymers -- Tactile information processing -- Electronic embedded systems; Computer integrated manufacturing systems -- Periodicals; Flexible manufacturing systems -- Periodicals; Mechatronics -- Periodicals; Productique -- Périodiques; Fabrication, Systèmes flexibles de -- Périodiques; Mécatronique -- Périodiques; Computer integrated manufacturing systems; Flexible manufacturing systems; Mechatronics; Periodicals; 629.89
• DOI: 10.1016/j.mechatronics.2015.04.001
• Languages: English
• ISSNs: 0957-4158

Abstract

Highlights: Main elements of e-skin systems are introduced. PVDF is the functional constituent of large and small area tactile sensor arrays. Machine Learning is used to extract meaningful information from raw tactile data. Advantages and limitations of ML-based tactile prediction systems are assessed. Requirements of the embedded electronic system are reviewed and discussed. Abstract: Due to its very peculiar features, the development of e-skin can be effectively tackled using a holistic approach. Starting from the definition of system specifications, the mechanical arrangement of the skin itself needs to be designed and fabricated together with the electronic embedded system, to move towards aspects such as tactile data processing algorithms and the communication channel interface. In this paper we present the design, the implementation and the results on the way of the development of an electronic skin (e-skin) system based on arrays of piezopolymer transducers. Focus of the paper is on both the development of innovative approaches for tactile information processing and electronic system embedding into the e-skin structure. In particular, Machine Learning technologies can provide a powerful tool to tackle the pattern-recognition problems involved in the tactile sensing framework and the ability of processing data represented as N-th order tensor is the key aspect of the presented research, which can be seen as an application of an existing method (Signoretto et al., 2011). The experimental session compares two different implementations of the ML-based framework, which differ in the learning paradigm adopted, namely SVM and ELM (K-ELM). The effectiveness of the adopted pattern-recognition technologies in the classification of touch modalities has been confirmed by addressing two different binary classification problems in an experiment involving 70 participants. The computational requirements for the hardware implementation of the proposed algorithm together with an overview of what exists in the existing literature are finally discussed.