Abstract:

Robot swarms that operate in a fully self-organized manner, without any central coordinating unit, have been widely demonstrated. These systems rely on decentralized architectures where collective behavior emerges from local interactions. This design provides key advantages such as scalability, fault tolerance, and the absence of single points of failure. However, it also introduces challenges in terms of system-level control and manageability.
In contrast, centralized systems are easier to design and control but lack scalability and are more vulnerable to failures at critical nodes.
In this talk, I will present a novel swarm architecture that enables self-organized hierarchy, combining the resilience of decentralized systems with the controllability of centralized ones. Using a heterogeneous team of ground and aerial robots, I will demonstrate how the swarm can self-organize into a dynamic hierarchical control structure through local asymmetric communication. I will present the results of experiments that illustrate capabilities such as autonomous sub-swarm splitting and merging, dynamic replacement of failed robots, and real-time adaptation of collective behavior, while preserving the key benefits of self-organization, including scalability and interchangeability of individual robots.

The extraction of features from speech is the most critical process in speech signal processing. Mel Frequency Cepstral Coefficients (MFCC) are the most widely used features in the majority of the speaker and speech recognition applications, as the filtering in this feature is similar to the filtering taking place in the human ear. But the current MFCC extraction is complex and requires time-frequency translations. Through our investigation, we were able to model a reservoir as a feature extractor capable of extracting the Mel Frequency Cepstral Coefficient (MFCC) without time-frequency domain translations. We have developed a real-time audio signal processing system by simplifying audio signal processing through the utilization of reservoir computers, which are significantly easier to train. We have established an experimental framework for end-to-end audio processing utilizing the reservoir and have investigated its capability to perform end-to-end audio signal processing.

Classical Cellular Automata (CCAs) are a powerful computational framework for modeling global spatio-temporal dynamics with local interactions. While CCAs have been applied across numerous scientific fields, identifying the local rule that governs observed dynamics remains a challenging task. Moreover, the underlying assumption of deterministic cell states often limits the applicability of CCAs to systems characterized by inherent uncertainty. This study, therefore, focuses on the identification of Cellular Automata on probability measures (CAMs), where cell states are represented by Bernoulli probability distributions and local rules act on these probabilistic configurations. This framework enables the modeling of systems with probabilistic uncertainty and spatially varying dynamics. Moreover, we formulate the local rule identification problem as a parameter estimation problem, and propose a meta-heuristic search based on Self-adaptive Differential Evolution (SaDE) to estimate local rule parameters accurately from the observed data. The efficacy of the proposed approach is demonstrated through local rule identification in multi-dimensional CAMs with varying neighborhood topologies and sizes.

We have created a new card game, Gakmoro, where two players each have seven cards numbered from 1 to 7. In each round, they secretly choose one to three cards to compete based on their total value; the first player to win two rounds wins the game. The unique feature of Gakmoro is that the cards chosen by the players must remain secret, and only the winner of each round is revealed. This secrecy is expected to add depth to the game by emphasizing the strategy of reading an opponent's hand. As designed, the game requires a dealer to secretly calculate the sum of the cards and announce only the winner. In this paper, instead of relying on a human dealer, we use card-based cryptography to virtually fulfill the dealer's role. In other words, we design an `unconventional' computation protocol that allows players to securely determine the winner without a dealer using a physical deck of cards.

Secure multi-party computation is an area in cryptography which studies how multiple parties can compare their private information without revealing it. Besides digital protocols, many unconventional protocols for secure multi-party computation using physical objects have also been developed. The vast majority of them use playing cards as the main tools. In 2024, Kaneko et al. introduced the use of a balance scale and coins in zero-knowledge proof protocols for pencil puzzles. In this paper, we extend the use of these tools to secure multi-party computation. In particular, we develop four protocols that can securely compute any n-variable Boolean function using a balance scale and coins.

UNO is a popular multiplayer card game. In each turn, a player has to play a card in their hand having the same number or color as the most recently played card. When having few people, adding virtual players to play the game can easily be done in UNO video games. However, this is a challenging task for physical UNO without computers. In this paper, we propose an unconventional protocol that can simulate virtual players using nothing but physical UNO cards. In particular, our protocol can uniformly select a valid card to play from each virtual player's hand at random, or report that none exists, without revealing the rest of its hand. The protocol can also be applied to simulate virtual players in other turn-based card or tile games where each player has to select a valid card or tile to play in each turn.

Thermal energy fuels random molecular motion which causes the molecules in a solution to mix and interact with each other. When interactions among those molecules, or subsets of them, allow them to bind together, self-assembly may occur. This process leads to the formation of intricate structures in nature, such as snowflakes, and also to increasingly complex structures in rationally designed, artificial self-assembling systems. In this talk, we discuss various methods by which molecular components composed of synthetic DNA can be designed so that, under the correct experimental conditions, the random thermal motion and interactions of those components yields the growth of structures that encode the results of targeted computations, which in turn can direct the shapes of the self-assembling structures. We show how these systems, fueled by randomness, are computationally universal in theory, and discuss limitations and errors that occur in practice and various methods for mitigating them that will hopefully lead to systems whose complexity eventually approaches the theoretical powers.

This paper establishes formal mathematical foundations linking Chaos Game Representations (CGR) of DNA sequences to their underlying k-mer frequencies. We prove that the Frequency CGR (FCGR) of order k is mathematically equivalent to a discretization of CGR at resolution 2^k x 2^k, and its vectorization corresponds to the k-mer frequencies of the sequence. Additionally, we characterize how symmetry transformations of CGR images correspond to specific nucleotide permutations in the originating sequences. Leveraging these insights, we introduce an algorithm that generates synthetic DNA sequences from prescribed k-mer distributions by constructing Eulerian paths on De Bruijn multigraphs. This
enables reconstruction of sequences matching target k-mer profiles with arbitrarily high precision, facilitating the creation of synthetic CGR images for applications such as data augmentation for machine learning-based taxonomic classification of DNA sequences. Numerical experiments validate the effectiveness of our method across both real genomic data and artificially sampled distributions. To our knowledge, this is the first comprehensive framework that unifies CGR geometry, k-mer statistics, and sequence reconstruction, offering new tools for genomic analysis and visualization. The web application implementing the reconstruction algorithm is available at https://tinyurl.com/kmer2cgr.

We investigate the computational power of Step-Cycle Chemical Reaction Networks (CRNs) when restricted to void reactions of size at most (3,1). Step-Cycle CRNs extend the previously introduced step CRN model by repeatedly cycling through a fixed sequence of species additions and reaction phases. We show that even under the severe constraint of trimolecular void rules --- which can only delete or preserve species --- the model retains full computational power. In particular, we prove that (3,1) void Step-Cycle CRNs can polynomially simulate (1) any general CRN, (2) any general Step CRN, and (3) any general Step-Cycle CRN. Ultimately, these results demonstrate that the Step-Cycle model retains its complete expressive power even when restricted to (3,1)-size void rules.

Graph-controlled insertion-deletion (GCID) systems are regulated extensions of insertion-deletion systems. In 2022, Alhazov et al. introduced time-varying systems as a restriction of GCID systems; here, the components are cyclically ordered. A rule is applied to a string in a component and the resultant string is moved to the next component as specified by this order. The language of the system is the set of all terminal strings collected in a distinguished component that is both the initial and the final one. With this restriction, here we obtain several new computational completeness results for some typical descriptional complexity measures as well-studied in the area of GCID systems.

Spiking Neural Networks (SNN) are models for "realistic" neuronal computation, which makes them somehow different in scope from "ordinary" deep-learning models widely used in AI platforms nowadays. SNNs focus on timed latency (and possibly probability) of neuronal reactive activation/response, more than numerical computation of filters.
So, an SNN model must provide modeling constructs for elementary neural bundles and then for synaptic connections to assemble them into compound data flow network patterns. These elements are to be parametric patterns, with latency and probability values instantiated on particular instances (while supposedly constant "at runtime"). Designers could also use different values to represent "tired" neurons, or ones impaired by external drugs, for instance.
One important challenge in such modeling is to study how compound models could meet global reaction requirements (in stochastic timing challenges), provided similar provisions on individual neural bundles. A temporal language of logic to express such assume/guarantee contracts is thus needed. This may lead to formal verification on medium-sized models and testing observations on large ones.
In the current article, we make preliminary progress at providing a simple model framework to express both elementary SNN neural bundles and their connecting constructs, which translates readily into both a model-checker and a simulator (both already existing and robust) to conduct experiments.

A common representation of crystal structures is by periodic graphs, i.e., graphs whose automorphism groups have subgroups of translational symmetries.
Such graphs may be represented as (finite) quotient graphs (called voltage graphs) whose edges are labeled by corresponding elements of their translational subgroup.
We outline a polynomial time algorithm for determining wether two finite bi-deterministically edge-labeled voltage graphs with voltages from corresponding translational subgroups generate isomorphic periodic graphs.
This algorithm aids in the classification of crystal structures and provides a method for determining whether two sets of building blocks can assemble isomorphic crystals.

Echo State Networks (ESNs) are stated in literature to use a random structure to project an input sequence into a higher dimensional space where the input becomes linearly separable. However, the linear mathematics used for this projection are incapable of increasing the dimensionality of the input, and the commonly used tanh activation function tends not to produce much nonlinearity. Therefore, any increase in dimensionality is due to the echoes of the ESN. We introduce Lagged Input Regression Computation to investigate what types of ESN can be replaced with simpler non-randomised structures. We show that tanh()-based ESNs behave as simple linear memory systems, whereas LeakyReLU provides a more effective non-linearity. We also show that the use of certain orthogonal polynomials in defining nonlinear memory capacity benchmarks gives a misleading impression of nonlinearity, due to the relevant high order polynomials nevertheless containing a linear term.

Physical learning methods have previously used the contrastive learning framework to alter the material properties of physical networks. This involves putting the networks into two distinct phases (the free phase and nudged phase), and applying an update rule inspired by Hebbian plasticity. In this paper, we relax the need for two distinct phases by exploiting a property of vibrational metamaterials called polycomputation, in which multiple functions can be computed simultaneously using the same parts. We also dispense with the assumption of a particular form for the update rule, and instead task an RNN that is shared across elements of a simulated metamaterial to map impressed forces into changes in the material properties of elements. This evolved RNN is meant to be a placeholder for the material properties that will be engineered into real elements when the metamaterial is transferred to reality, because it represents a transformation of impressed forces into altered properties. Such transformations are already used in the elements of existing metamaterials, taking the form of rules like "if heated, expand'' and "if cooled, contract''. Here we show proof-of-concept that our back-and-forth algorithm can execute forward outputs and backward error correction simultaneously to match arbitrary target outputs.