My project: Multi-transduction neuromorphic skin READ MORE
Research Topics: Neuromorphic Engineering
Institution: Istituto Italiano di Tecnologia
I have always enjoyed gaining a deeper understanding of contexts, especially seeing something you should have learned in school in the real world and finally just getting it. I have had several experiences like that in the topics of sensing and neuromorphic engineering. So I am happy to now combine both.
Is special, because the research topics, from various fields, are so interwoven with each other. Thus we will get to collaborate early on and to gain deeper insight into the various topics. Thus, we have the chance to early on weave a network with researchers from different backgrounds (both cultural and professional), transfer our skills and overall strongly profit from each other.
Development of a novel artificial sensitive skin based on different physical transduction technologies of the tactile physical contact with spike-based neural encoding. The addressed physical transduction technologies will be capacitive and piezoelectric in particular piezoelectric polymeric films like PVDF-TrFe. We will exploit complementary features of the two transduction principles i.e. capacitive transducers efficiently measure contact phenomena in the low frequency range (from DC to up to some tenths of Hz); on the other hand, piezoelectric transducers cover the higher frequency band (from some Hz up to 1 kHz). The combination of the two can span over the entire frequency range of human tactile transduction (from DC up to 1 kHz). We will develop spike-based readout based on the neural encoding mechanisms studied in WP1 and WP2 and design a new spiking skin with interleaved capacitive and piezoelectric sensors. We will develop circuit architectures to efficiently encode the physical contact information into spike trains of the proper frequency. We will develop array geometries of the spiking neurons in such a way as to efficiently couple the two transductions. Effective spike train frequency encoding coupled with smart integration of information from spiking neurons with different transduction will be developed.
Novel spiking neuron circuit architectures for effectively encode the physical contact information; arrays of spiking neurons for different body regions (e.g. different geometrical pitch and neuron size). The neural network hardware will be implemented with dedicated CMOS microelectronics circuits with post-processing steps.