Analysis design, and simulation of menristor-based commercial crossbar architectures

This work is framed within the field of Artificial Intelligence (AI), specifically in one of the fields with the greatest future projection: neuromorphic systems. These systems try to emulate the functioning of the human brain by means of neural impulse networks, for whose implementation different ways are being explored at hardware level. Among the proposed devices is the memristor, a component theorized by Leon Chua in the 1970s, whose fabrication was not achieved until 2008 at HP Labs. Its main attraction lies in the reduction of power consumption and latency associated with data transfer between the memory system and the processing system, which are considered the main bottlenecks in AI systems. This work focuses on the study, characterization, and simulation of typical struc-tures integrating memristors. First, the field of artificial intelligence in which the work is situated is introduced, followed by a deeper understanding of the theoretical framework of neuromorphic systems, memristors, and their most common topologies. A chapter is dedicated to the characterization of the technology used, including both the transistor and the memristor. The bulk of the work is oriented to the analysis of a type of problem little treated in the literature. Finally, several simulations using the crossbar structure are performed.