Ultrafast Hybrid Computing Systems Enabled by Memristor‐Based Quadratic Programming Circuits

Implementing algorithms purely on digital computing platforms dramatically halts the performance of conventional computing systems. Revolutionary computing systems with extreme energy efficiency and high accuracy are demanded to handle the growing computing tasks. Here, the research on hybrid analog–digital computing platforms enabled by memristor‐based optimization solvers for achieving ultrafast computations is presented. By utilizing tunable memristors as parameters to solve linear programming (LP) and quadratic programming (QP) problems, a real‐time control algorithm for micro air vehicles (MAVs) and a support vector machine (SVM) algorithm for cancer diagnosis are implemented. These experiments demonstrate over 2000x speed‐up compared to conventional digital platforms, with negligible energy consumption, using a memristor‐based system consisting of six memristors. These findings underscore the vast potential of memristor‐based optimization solvers not only in hybrid analog–digital computing platforms but also as a transformative solution for a wide range of modern computing challenges. This approach promises significant advancements in energy efficiency and ultrafast speed, positioning it as a leading contender for next‐generation computing paradigms. This article introduces a computing system enabled by memristor‐based quadratic programming (QP) circuit. It is the first demonstration of a memristor‐based QP solver that solves constrained QP problems in a small form‐factor silicon chip, which can be integrated into any micro‐computing system with limited energy, power, and space. This computing system achieves unparalleled speed (2000x improvement), energy efficiency (2.9x improvement), and adaptability, showcasing its capability in control and edge computing.