Probabilistic Autonomous Data Acquisition Using Stochastic MTJ Based p-Bits

Spintronic devices that utilize thermal noise for producing a natural stochastic response have enabled the concept of probabilistic-bits (p-bits), opening the door for a spectrum of untapped opportunities and new applications. Here we propose a probabilistic autonomous data acquisition approach that translates the stochastic response of a p-bit into a cognitive response of deciding whether to sample data or not. In the proposed approach, a modified p-bit design with enhanced tunability is used and is driven by an analog feature extraction circuit. Simulation results on data from an active seismic survey demonstrate the feasibility of lossless probabilistic data acquisition, with close to 50 % saving in the active operation time of the system and in the number of generated samples.