Heidemarie Krüger and her team are developing memristor-based components aimed at setting new standards in energy efficiency and computing power. This real-time, resource-efficient technology could support applications like autonomous vehicles and industrial systems.
The core of this innovation lies in memristors, which function similarly to synapses in the brain. They not only store information but also process it simultaneously, drastically reducing energy losses and enabling fast, decentralized data analysis.
Unlike conventional computers that operate on binary data, memristors can handle continuous states, allowing for advanced algorithms that mimic neural networks. This flexibility opens doors for predictive maintenance and real-time analysis in critical fields such as autonomous driving.
The development began with a serendipitous laboratory discovery, revealing the hysteretic memristance behavior essential for the device's memory capabilities. This led to the creation of artificial synapses from a bismuth-iron oxide combination, supported by significant funding from the Federal Agency for Jump Innovations.
Initial pilot projects in collaboration with the Technical University of Freiberg have demonstrated the chip's ability to detect minute changes and accurately predict wear patterns, showcasing its potential in edge computing by processing data locally for enhanced security and independence.
As traditional processors struggle with increasing data demands, this neuromorphic approach combines memory and processing units, significantly reducing energy consumption and expanding the potential for AI systems. The current prototype features 32 memristors, with plans to scale to over 200 for more complex neural networks.