Time-domain readout cuts noise errors in dense tactile sensor arrays for robots

Taro Tako, Hiroshi Kaminaga and Taisuke Kobayashi have described, in an arXiv paper, a new way to read cross-point resistive arrays, which suit dense touch sensing on robots because they need less wiring. Usual frequency-based readout can go wrong in robots, where long cables and noise distort the signals.

Their method fits a time-domain model to the whole measured waveform instead of relying only on spectral amplitudes. In simulations of a 16 x 16 array, it cut the mean readout error by 20%, and over 1,000 trials it lowered the mean error by 24.6% and its standard deviation by 77.6% compared with the conventional method.