Collective motion of microscopic oscillators enables a new form of physical computing: Researchers from Konstanz and Stuttgart bring “reservoir computing” to life.

Researchers at the Universities of Konstanz and Stuttgart have demonstrated a new approach to computation: a computer made from a few hundred microscopic particles oscillating in a liquid. Instead of processing information through carefully fabricated electronic circuits, their system performs computations by harnessing the complex collective dynamics of interacting particles.

Conventional computers rely on vast numbers of simple transistors that switch in precisely controlled ways. Even the artificial neural networks that power large language models consist of many basic units whose connections must be carefully designed and optimized. As tasks grow more complex, this approach demands increasing hardware resources and energy consumption.

Researchers from Konstanz led by Clemens Bechinger take a fundamentally different approach: Instead of processing data with digital circuits, they drive arrays of several hundred oscillating microscopic particles with the data as an input signal. These particles are immersed in a liquid, which couples the data-driven motion between individual particles. In this way, the systems respond to input data with complex collective dynamics which is largely influenced by the liquid flows set in motion by the particles.

The complex physical dynamics of the microparticle arrays thus naturally produce complex representations of the input data, which – in traditional computers – would require costly operations. “Based on such complex data-representations, computation becomes remarkably simple: Selected features of the microparticle dynamics are measured and can then be combined to yield desired outputs”, says first author Veit-Lorenz Heuthe. This approach is known as reservoir computing and the researchers from Konstanz and Stuttgart were the first to demonstrate this novel computing paradigm in a microscopic many-particle system.

“Crucially, the reservoir dynamics does not need to be fully understood, in stark contrast to conventional computing, where every operation must be precisely controlled. As long as the reservoir responds reliably, its physics can be directly harnessed for computation”, remarks Clemens Bechinger, a professor for soft condensed matter at the University of Konstanz and member of the Centre for the Advanced Study of Collective Behaviour.
Using only fluid-coupled microparticle arrays as reservoirs, the researchers were able to predict chaotic time series with high accuracy. Even more strikingly, the reservoir enabled the detection of extremely subtle anomalies in the input data. This capability is particularly relevant for noisy real-world data, such as seismic measurements or climate records, where subtle shifts in the underlying dynamics can be early warning signs of catastrophic events.

Although the current demonstration serves as a controllable laboratory model system only, it shows that useful computation can emerge from the collective motion of interacting microscopic particles. In this way, the study opens new perspectives for energy-efficient computing and intelligent edge-level sensing devices in which complex physical systems themselves perform information processing.

Original publication:
Veit-Lorenz Heuthe, Lukas Seemann, Samuel Tovey und Clemens Bechinger
Reservoir computing from collective dynamics of active colloidal oscillators. Commun. AI Comput. 1, 6 (2026).
DOI: https://doi.org/10.1038/s44488-026-00001-3