Ferdinand Peper’s research while affiliated with National Institute of Information and Communications Technology and other places

Brownian computers utilize thermal fluctuations as a resource for computation and hold promise for achieving ultra-low-energy computations. However, the lack of a statistical direction in Brownian motion necessitates the incorporation of ratchets that facilitate the speeding up and completion of computations in Brownian computers. To make the ratchet mechanism work effectively, an external field is required to overcome thermal fluctuations, which has the drawback of increasing energy consumption. As a remedy for this drawback, we introduce a new approach based on one-dimensional (1D) quantum Brownian motion, which exhibits intrinsic unidirectional transport even in the absence of external forces or asymmetric potential gradients, thereby functioning as an effective pseudo-ratchet. Specifically, we exploit that quantum resonance effects in 1D systems divide the momentum space of particles into subspaces. These subspaces have no momentum inversion symmetry, resulting in the natural emergence of unidirectional flow. We analyze this pseudo-ratchet mechanism without energy dissipation from an entropic perspective and show that it remains consistent with the second law of thermodynamics.

The Internet of Things (IoT) has revolutionized the way we interact with everyday objects by connecting sensors/actuators to the Internet to monitor and control various aspects of our environment. As the number of IoT devices continues to grow, efficient medium access control protocols are needed to ensure that communication takes place in a reliable and scalable manner. In this paper, we investigate the performance of the Asynchronous Pulse Code Multiple Access (APCMA) protocol in massive IoT scenarios through simulations and analysis. We show that APCMA is able to outperform CSMA/CA in terms of the number of successfully transmitted messages in scenarios with up to 30,000 transmitting sensor nodes while also being able to utilize the channel more efficiently. We also examine the sensitivity of APCMA’s performance with respect to its parameters.

We analyze the thermodynamic cost of a logically reversible Brownian Turing machine operating in the first-passage time protocol based on the stochastic thermodynamics of resetting. In this framework, the thermodynamic cost of computation is the reset entropy production, which is interpreted as the information reduction by a resetter external to the computer. At the level of a single trajectory, the reset entropy production is associated with unidirectional transitions and is a function of the time-dependent distribution probability. We analyze an approximation that replaces the distribution probability with the empirical sojourn time, which can be obtained at the single-trajectory level. The approximation is suitable for the numerical analysis by the Gillespie algorithm and provides a reasonable average value for the reset entropy.

Brownian circuits use the fluctuations of signals, implemented by tokens, to drive computation. They have been shown to significantly reduce the required complexity of circuits or platforms implementing these circuits, such as cellular automata. The original model of Brownian circuits eyed computation models in which operations on individual tokens are at the core. This paper discusses models in which collections of tokens are used as signals in Brownian circuits, in particular neural networks. We show how previously proposed Brownian circuit primitives can be employed to implement neural functionality, like thresholding, synchronization, and learning.KeywordsBrownian circuitsFluctuation-driven computationAsynchronous circuitsNeural networks

We analyze the thermodynamic cost of a logically reversible Brownian Turing machine operating in the first-passage time protocol based on the stochastic thermodynamics of resetting. In this framework, the thermodynamic cost of computation is the reset entropy production, which is interpreted as the information reduction by a resetter external to the computer. At the level of a single trajectory, the reset entropy production is associated with unidirectional transitions and is a function of the time-dependent distribution probability. We analyze an approximation that replaces the distribution probability with the empirical sojourn time, which can be obtained at the single-trajectory level. The approximation is suitable for the numerical analysis by the Gillespie algorithm and provides a reasonable average value for the reset entropy.

Asynchronous Pulse Code Multiple Access (APCMA) is a pulse-based communication system for massive IoT. It enables high-density communication because it can decode messages with a high probability even if they collide. We evaluate the performance of APCMA with Chirp Spread Spectrum-modulated pulses (CSS-APCMA) by comparing it with On-Off Keying APCMA (OOK-APCMA) and LoRa. Our first experiment shows that APCMA has a lower packet error rate (PER) than LoRa in a congested environment. Our second experiment shows that CSS-APCMA has a smaller PER than OOK-APCMA in a noisy environment. Our results confirm that CSS-APCMA is suitable for massive IoT.

We analyze a token-based Brownian circuit in which Brownian particles, coined `tokens,' move randomly by exploiting thermal fluctuations, searching for a path in multi-token state space corresponding to the solution of a given problem. The circuit can evaluate a Boolean function with a unique solution. However, its computation time varies with each run. We numerically calculate the probability distributions of Brownian adders' computation time, given by the first-passage time, and analyze the thermodynamic uncertainty relation and the thermodynamic cost based on stochastic thermodynamics. The computation can be completed in finite time without environment entropy production, i.e., without wasting heat to the environment. The thermodynamics cost is paid through error-free output detection and the resets of computation cycles. The signal-to-noise ratio quantifies the computation time's predictability, and it is well estimated by the mixed bound, which is approximated by the square root of the number of token detections. The thermodynamic cost tends to play a minor role in token-based Brownian circuits in computation cycles. This contrasts with the logically reversible Brownian Turing machine, in which the entropy production increases logarithmically with the size of the state space, and thus worsens the mixed bound.