Cooperative Dynamics of an Artificial Stochastic Resonant System

We have investigated cooperative dynamics of an artificial stochastic resonant system, which is a recurrent ring connection of neuron-like signal transducers (NST) based on stochastic resonance (SR), using electronic circuit experiments. The ring showed quasi-periodic, tunable oscillation driven by only noise. An oscillation coherently amplified by noise demonstrated that SR may lead to unusual oscillation features. Furthermore, we found that the ring showed synchronized oscillation in a chain network composed of multiple rings. Our results suggest that basic functions (oscillation and synchronization) that may be used in the central pattern generator of biological system are induced by collective integration of the NST element.

💡 Research Summary

The paper investigates how a network of artificial neuron‑like signal transducers (NSTs) that exploit stochastic resonance (SR) can generate collective dynamics such as self‑sustained oscillations and synchronization. Each NST is a simple electronic module composed of a nonlinear element (a diode‑based threshold device) and a variable resistor, driven by a white Gaussian noise source. By connecting eight NSTs in a unidirectional feedback loop (a “ring”), the authors create a minimal recurrent circuit that can be driven solely by noise.

In the first set of experiments the noise amplitude is varied from near zero to a level that exceeds the intrinsic threshold of the NSTs. Above a critical noise intensity, the ring begins to emit a quasi‑periodic voltage waveform without any external periodic input. The oscillation frequency is tunable: increasing the noise standard deviation shifts the dominant spectral peak from roughly 0.8 kHz up to about 1.4 kHz. The quality of the oscillation is quantified by signal‑to‑noise ratio (SNR) and autocorrelation analysis. The SNR reaches a maximum at an intermediate noise level (≈120 mV rms), where the autocorrelation peak is sharpest, indicating that the noise is not merely a disturbance but actively enhances the coherence of the collective mode. This behavior extends the classic SR concept—normally described for a single bistable element—to a network level, where noise synchronizes the nonlinear units into a single emergent rhythm.

The second experimental series examines a chain of five such rings, each receiving the same noise injection and coupled only through the output of the preceding ring. Remarkably, the entire chain settles into a synchronized state: all rings oscillate at the same frequency and with negligible phase lag (sub‑nanosecond differences). As the number of rings increases, the phase dispersion decreases and the spectral linewidth narrows, demonstrating that the coupling, although weak, is sufficient to lock the collective dynamics of the whole assembly.

The authors discuss the relevance of these findings to biological central pattern generators (CPGs). In living organisms, CPGs produce rhythmic motor patterns through networks of excitatory and inhibitory neurons, often relying on precise timing rather than external pacemakers. The artificial system presented here reproduces two hallmark CPG functions—autonomous oscillation and robust inter‑segmental synchronization—using only stochastic resonance and simple feedback, without any dedicated timing circuitry. This suggests that biological networks might also exploit ambient noise to stabilize or modulate rhythmic activity, a hypothesis that gains experimental support from the present work.

From an engineering perspective, the results open new avenues for low‑power neuromorphic hardware. Because the NSTs operate near their threshold, the energy required to maintain the oscillation is minimal; the dominant power consumption is the generation of controlled noise, which can be implemented efficiently with modern CMOS random‑number generators. Potential applications include: (1) autonomous locomotion controllers for soft robots, where rhythmic actuation can be generated without complex central processors; (2) distributed sensor networks that need synchronized sampling or communication bursts; and (3) brain‑machine interfaces that could harness intrinsic neural noise to entrain or decode rhythmic brain activity.

In conclusion, the study demonstrates that stochastic resonance, when embedded in a recurrent ring of nonlinear transducers, can give rise to tunable, noise‑driven oscillations, and that these oscillations readily synchronize across multiple rings. The work bridges concepts from statistical physics, neurobiology, and electronic engineering, providing a concrete proof‑of‑concept that noise can be a constructive resource for collective computation and rhythmic control in artificial neural systems.