Development of a Thermo-Computing Platform

There is an increasing demand for the recognition of context from sensor data. This is presently achieved through running complicated statistical signal processing algorithms with significant computing and memory requirements. In order to reduce the complexity and power requirements, unconventional computing platforms are being considered, which rely on the responses of the materials or devices instead of digitizing information and processing them. Herein, for the first time to the best of our knowledge, we propose a thermo-computing platform, which can shift much of the complex computations to the sensors. The proposed platform employs an entirely passive network of thermistors for processing temporal data. We present results that confirm the capability of the thermo-computer in processing data. A thermo-computer was then used for processing benchmark data, and its results are compared against algorithmic programming. The proposed platform, in addition to its use as a thermal computer, can lay the foundation for the development of cognizant sensor that utilize thermistor-like devices, such as MOX multi-gas sensors.