Design and application of physical contextual computers

The widespread adoption of Internet of Things (IoT) devices has ushered in an era of extensive sensor-generated data, leading to the need for improved communication capabilities, data storage, and energy efficiency. A significant proportion of the power used by these devices, ranging from 60% to 98%, is dedicated to energy dissipation through communication lines. This thesis addresses these requirements through innovative computing approaches that center on processing data in close proximity to sensors. A key contribution of this research is the introduction of "thermo-computing," a ground breaking concept that uses the unique characteristics of materials and devices to process data over time. Thermo-computing leverages an entirely passive network of thermistors for data manipulation, promising significant energy savings. This work, supported by extensive experiments and careful analysis, firmly establishes thermocomputing as a transformative method for data processing. Additionally, the thesis examines the capabilities of 3D-printed computing platforms for real-time data processing and the classification of sensory data. It investigates how memory, nonlinearity, and sampling rates affect the performance of these processors, highlighting their cost-effectiveness and ease of integration into existing smart and 3D-printed intelligent systems, marking a noteworthy advancement in the domain of 3D-printed systems. Furthermore, the research bridges the gap between software-based reservoir computing and its physical counterpart, introducing a novel reservoir computing platform. This pioneering platform enhances our understanding of the applications of physical reservoir computing and explores the transfer of energy within physical systems acting as reservoirs inspired by neural networks. This innovative approach has the potential to transform physical computing platforms and offer new solutions across areas such as networks, sensing, and computing.