ABSTRACT In Quantum chemical computation, numerical schemes such as the Hartree–Fock (HF) and density functional theory (DFT) are widely used to solve the Schrödinger equation numerically, to realize experiment‐free prediction and analysis of key molecular properties such as structure and energy. Computing one‐electron integrals, such as kinetic energy integrals and nuclear attraction integrals, is essential in both HF and DFT to characterize the molecular electronic states. However, as molecules of practical interest grow in size and angular momentum, computing one‐electron orbitals becomes computationally expensive in most cases. Although computing kinetic energy integrals on CPUs is straightforward, bottlenecks in CPU‐GPU data transfer have often been overlooked. In this study, we propose an efficient method to compute both the kinetic‐energy and nuclear‐attractive integrals on GPUs. First, we explicitly and symbolically expand recurrence relations based on the Obara–Saika and McMurchie–Davidson methods to eliminate redundant operations, thus improving computational efficiency. Second, we implemented a hybrid method that selects the best/fastest of both methods depending on the integration task. Third, we achieved further speedups by using CUDA streams to parallelize the execution of multiple kernels and efficiently utilize multiprocessor resources on the GPU. Computational experiments using NVIDIA A100 GPUs and Intel Xeon Gold 6338 CPU on relevant molecules of interest demonstrated the superiority of our one‐electron integral GPU implementations, achieving a speedup of 20.2 times over PySCF, and a speedup of 132.6 times over GPU4PySCF.

Searching for activation functions has the potential to tackle generalization across different problem domains. In this paper, we study the Extreme Learning Machine with Learnable Activation Functions (ELMA), leveraging ordered trees and bilevel optimization via Differential Evolution algorithms. We analyze the effectiveness in learning function surrogates for the inverse inference of gains and parameters in high-performance machine control. Our computational experiments, spanning diverse machine control scenarios including motor position control, motor velocity control, crane stabilization, inverted pendulum, and magnetic levitation demonstrate that ELMA achieves superior learning and generalization performance compared to other existing frameworks.

ABSTRACT Motion planning for multiple robots operating in constrained environments is a fundamental challenge in robotics. Overcoming limitations in scalability and solution quality in existing multi‐robot planners is essential for safe, efficient, and collision‐free navigation. Contemporary multi‐robot coordinated planning is constrained by the granularity of probabilistic roadmap configurations: Rendering feasible probabilistic roadmaps that can be used effectively and efficiently by sample‐based path‐planning schemes, such as dRRT*, is often time‐consuming. To address roadmap generation and effective sampling in coordinated motion planning for mobile robots, we introduce RADES (Rank‐based Differential Evolution with a Successful Archive), a novel gradient‐free optimization algorithm for multi‐robot coordinated planning. RADES enhances sampling in lattice‐based roadmap configurations by integrating rank‐based selection, successful‐mutation archiving, and stagnation‐control mechanisms. Comprehensive computational experiments across intersection scenarios involving up to 12 robots demonstrate that RADES outperforms seven established gradient‐free optimization techniques and two state‐of‐the‐art winners from CEC 2024 in terms of solution cost and convergence performance. Our approach facilitates the use of gradient‐free optimization algorithms to sample the search space of feasible and safe multi‐robot roadmaps.

ABSTRACT In computational quantum chemistry, the computation of three‐center two‐electron repulsion integrals (also termed three‐center ERIs) is essential for density fitting. Due to the large number of integral elements and the induced combinatorial computational complexity, the community has actively pursued the acceleration/speedup of ERI calculations to achieve pragmatic levels of efficiency. From the perspective of GPU acceleration, atomicAdd is known to incur significant memory overhead: The frequent collisions and retrials of value aggregation in global GPU memory lead to substantial performance degradation. To tackle this issue, we propose new thread mapping strategies for three‐center two‐electron integrals on GPUs, aiming at reducing the computational cost associated with value aggregation. Our methods are based on the idea of suitable substitutions of device‐level reduction ( atomicAdd ) with efficient warp‐ and thread‐level reduction, such as warp‐shuffle and register accumulation. As a result, our computational experiments using an Intel Xeon Gold 6338 CPU, an NVIDIA A100 GPU, and relevant molecules of interest show the superiority against the conventional thread mapping scheme, achieving up to 2.76 speedups to compute three‐center ERIs more efficiently. Moreover, compared to well‐known quantum chemistry software such as PySCF and GPU4PySCF, our method achieved up to speedups over PySCF and up to speedups over GPU4PySCF. Our method has the potential to further enhance the performance, extensibility, and versatility of GPU‐accelerated quantum chemical computations.

The safety and accuracy of robotic navigation hold paramount importance, especially in the realm of soft continuum robotics, where the limitations of traditional rigid sensors become evident. Encoders, piezoresistive, and potentiometer sensors often fail to integrate well with the flexible nature of these robots, adding unwanted bulk and rigidity. To overcome these hurdles, our study presents a new approach to shape sensing in soft continuum robots through the use of soft e-textile resistive sensors. This sensor, designed to flawlessly integrate with the robot’s structure, utilizes a resistive material that adjusts its resistance in response to the robot’s movements and deformations. This adjustment facilitates the capture of multidimensional force measurements across the soft sensor layers. A deep Convolutional Neural Network (CNN) is employed to decode the sensor signals, enabling precise estimation of the robot’s shape configuration based on the detailed data from the e-textile sensor. Our research investigates the efficacy of this e-textile sensor in determining the curvature parameters of soft continuum robots. The findings are encouraging, showing that the soft e-textile sensor not only matches but potentially exceeds the capabilities of traditional rigid sensors in terms of shape sensing and estimation. This advancement significantly boosts the safety and efficiency of robotic navigation systems.

Socially aware robotic applications such as companion and therapeutic robots usually require human emotions or intent to be conveyed. As the scope of these applications increases, the need for recognizing affective touch gestures which are often used to convey these emotions or intent becomes eminent. However, existing touch gesture recognition modalities either have low recognition accuracy or depend heavily on carefully hand-crafted features, therefore limiting their deployment in real-life applications. Motivated by the need for learning models with superior accuracy which do not rely on manually selected hand-crafted features, this paper proposes an optimized class-weighted CNN-LSTM for social touch gesture recognition evaluated on the CoST and HAART datasets. Specifically, contrary to vanilla training schemes where equal importance is given to each class in the dataset, different class weights are introduced to give priority to classes that are difficult for the network to distinguish during training. Furthermore, the weights associated with each of the classes are obtained through optimization using Genetic Algorithm. The proposed model demonstrates superior performance compared with other existing models in the literature.

The quest for the efficient adaptation of multilegged robotic systems to changing conditions is expected to render new insights into robotic control and locomotion. In this paper, we study the performance frontiers of the enumerative (factorial) encoding of hexapod gaits for fast recovery to conditions of leg failures. Our computational studies using five nature-inspired gradient-free optimization heuristics have shown that it is possible to render feasible recovery gait strategies that achieve minimal deviation to desired locomotion directives with a few evaluations (trials). For instance, it is possible to generate viable recovery gait strategies reaching 2.5 cm, (10 cm.) deviation on average with respect to a commanded direction with 40 – 60 (20) evaluations/trials. Our results are the potential to enable efficient adaptation to new conditions and to explore further the canonical representations for adaptation in robotic locomotion problems.

Lattice paths are functional entities that model efficient navigation in discrete/grid maps. This paper presents a new scheme to generate collision-free lattice paths with utmost efficiency using the bijective property to rooted ordered trees, rendering a one-dimensional search problem. Our computational studies using ten state-of-the-art and relevant nature-inspired swarm heuristics in navigation scenarios with obstacles with convex and non-convex geometry show the practical feasibility and efficiency in rendering collision-free lattice paths. We believe our scheme may find use in devising fast algorithms for planning and combinatorial optimization in discrete maps.

Due to the increased demand for music streaming/recommender services and the recent developments of music information retrieval frameworks, Music Genre Classification (MGC) has attracted the community's attention. However, convolutional-based approaches are known to lack the ability to efficiently encode and localize temporal features. In this paper, we study the broadcast-based neural networks aiming to improve the localization and generalizability under a small set of parameters (about 180k) and investigate twelve variants of broadcast networks discussing the effect of block configuration, pooling method, activation function, normalization mechanism, label smoothing, channel interdependency, LSTM block inclusion, and variants of inception schemes. Our computational experiments using relevant datasets such as GTZAN, Extended Ballroom, HOMBURG, and Free Music Archive (FMA) show the state-of-the-art classification accuracies in MGC. Our approach offers insights and the potential to enable compact and generalizable broadcast networks for music classification.

Proportional-integral-derivative (PID) control is ubiquitous in industrial automation tasks, and the parameter tuning of the gains is challenging due to nonlinearity and stagnation in local optima. In this paper we present a differential particle scheme based on stagnation-based selection mechanism, and evaluate its effectiveness in the stabilization of a nonlinear inverted pendulum and a magnetic levitation system. Our computational experiments show the feasibility to avoid stagnation, the lower variability of convergence over independent runs, and the feasibility to converge to significantly better fitness values compared to relevant heuristics in the literature. We believe our approach offers the building blocks to build stagnation-free nature inspired optimization algorithms useful for adaptive control and tuning.

Addressing the subset sum problem is relevant to study resource management problems efficiently. In this paper, we study a new scheme to sample solutions for the subset sum problem based on swarm-based optimization algorithms with distinct forms of selection pressure, the balance of exploration-exploitation, the multimodality considerations, and a search space defined by numbers associated with subsets of fixed size. Our experiments show that it is feasible to find optimal subsets with few number of fitness evaluations, and that Particle Swarm Optimization with Fitness Euclidean Ratio converges faster to the global optima with zero variability over independent runs. Since the search space is one-dimensional and friendly to parallelization schemes, our work is potential to study further classes of combinatorial problems using swarm-based optimization algorithms and the representation based on numbers.

Motion planning approaches aided by learning schemes have achieved relevant results in the community, particularly in terms of rendering new paths efficiently and adapting to new environments/situations through encoder-decoder frameworks and latent space configurations. This paper evaluates the feasibility of learning motion planning functions for robot manipulators using a linear transition of the configuration space. Our computational experiments involving a relevant set of learning architectures have shown the feasibility and the efficiency in finding motion planning functions that meet user-defined criteria. Our approach contributes to realizing the practical efficiency to tackle the learning-based motion planning problem. Due to the amenability to parallelization schemes, our approach is potential to tackle larger degrees of freedom.

Path planning with smoothness considerations is of relevant interest to ensure the safety and the comfortability of passengers in mobile and vehicle navigation. In this paper, we present our preliminary results in computing smooth paths from observed mobile robot trajectories. Our approach enables the generation of alternative paths safer paths for navigation, and is potential to extend towards the fitting and fairing of curves with utmost efficiency.

The estimation of grasp states in myoelectric prosthetic hands is relevant for ergonomic interfacing, control and rehabilitation initiatives. In this paper we evaluate the possibility to infer the grasp state of a prosthetic hand from RGB frames by using well-known deep learning architectures in testing scenarios involving variations of brightness, contrast and flips. Our results show the feasibility, the attractive accuracy and efficiency to estimate prosthetic hand poses with a GoogLeNet-based deep architecture using relatively few training frames.

The study of succinct representation of graphs has received relevant attention to allow efficiency in modeling interconnected systems. Related work on graph representation has achieved compact encodings by benefiting from structural regularities, such as triangularity, separability, planarity, symmetry and sparsity; whereas the case of arbitrary unstructured graphs has remained elusive. In this paper, we propose an scheme to represent arbitrary graphs by using the smallest integer number. We believe our approach is useful to represent graphs efficiently.

Modular Graphs are relevant mechanisms to represent large-scale and hierarchical relationships among entities. In this paper we propose mechanisms to allow the enumerative representation of modular graphs. We believe our approach is useful to represent systems with modular structures, such as mechanical systems, succinctly and canonically.

Identifying granular patterns of differentiation and learning predictors of product performance are key drivers to capitalize on competitive market segments. In this paper, we propose an approach to identify granular product patterns by using Hierarchical Clustering, and to learn predictors of product performance from historical data by using Genetic Programming. Computational experiments using more than twenty thousand vehicle models collected over the last thirty years shows (1) the feasibility to identify vehicle differentiation at different levels of granularity by hierarchical clustering, and (2) the good predictive ability of learned fuel consumption predictors in vehicle cluster. We believe our approach introduces the building blocks to further advance on studies regarding product differentiation and market segmentation by using data-intensive approaches.

Optimal topologies in networked systems is of relevant interest to integrate and coordinate multi-agency. Our interest in this paper is to compute the root location and the topology of minimal-length tree layouts given n nodes in a polygonal map, assuming an n-star network topology. Computational experiments involving 600 minimal tree planning scenarios show the feasibility and efficiency of the proposed approach.

Combinations of m out of n are ubiquitous to model a wide class of combinatorial problems. For an ordered sequence of combinations, the unranking function generates the combination associated to an integer number in the ordered sequence. In this paper, we present a new method for unranking combinations by using a gradient-based optimization approach. Exhaustive experiments within computable allowable limits confirmed the feasibility and efficiency of our proposed approach. Particularly, our algorithmic realization aided by a Graphics Processing Unit (GPU) was able to generate arbitrary combinations within 0.571 seconds and 8 iterations in the worst case scenario, for n up to 1000 and m up to 100. Also, the performance and efficiency to generate combinations are independent of n, being meritorious when n is very large compared to m, or when n is time-varying. Furthermore, the number of required iterations to generate the combinations by the gradient-based optimization decreases with m in average, implying the attractive scalability in terms of m. Our proposed approach offers the building blocks to enable the succinct modeling and the efficient optimization of combinatorial structures.

Being ubiquitously used as anti-adhesive and wound-covering mechanisms, thin films have potential therapeutic uses as cell sheets to target inner organs while navigating narrow environments. A significant challenge to realize versatile films lies in achieving compact storage and efficient transport while ensuring coherency in curvature-bounded environments. In this paper, we propose a folding mechanism of a curved film by using a spiral approach, enabling efficient unfolding and flexible plasters with curved surfaces. Our experiments using gelatin-based films with curved surfaces shows the superior indwelling ability in terms of chromaticity level compared to the conventional planar films, as well as the efficient unfolding in the order of seconds. Our results presents the theoretical and experimental building blocks to realize a versatile class of films which are able to navigate narrow environments, and unfold efficiently and flexibly.

Graphs with self-loops enable to represent a large variety of interactions in natural and artificial systems, allowing not only inter-connectivity among heterogeneous entities but also the self-dependence of entities, e.g. the recursive and autonomous nature of dynamical systems. In this paper we present new bijective constructs which enable the numerical representation of graphs with self loops (or loopy graphs). In particular, we study the case of (1) undirected and (2) directed graphs with n nodes and m edges with self-loops. Our proposed approach realizes the succinct representations by using integer numbers in which rigorous computational experiments show the efficiency of our proposed algorithms: the complexity follows a quasi-linear behaviour as a function of the number of edges (which is independent of the number of nodes). Furthermore, as direct consequence of our constructs, we propose list structures having O(m) space complexity, which realize the linear space complexity depending only on the number of edges (the list is independent of n). We believe that our bijective algorithms are useful to tackle problems involving sampling of graphical models, network design as well as process planning by using number theory and sample-based learning.

This paper aims at computing minimal-length tree layouts given an n-star graph in a polygonal map. This problem is strongly related to the edge bundling problem, which consists of compounding the edges of an input graph to obtain topologically compact graph layouts being free of clutter and easy to visualize. Computational experiments using a diverse set of polygonal maps and number of edges in the input graph shows the feasibility, efficiency and robustness of our approach.

Graphs denote useful dependencies among objects ubiquitously. This paper introduces new and simple bijections to the integer grid to enable the succinct, canonical and efficient representations of labeled graphs; whereas previous work has focused on regularities in structure such as triangularity, separability, planarity, symmetry and sparsity. By succinct we imply that space is information-theoretically optimal, by canonical we imply that generation of instances is unique, and by efficient we imply that coding and decoding take polynomial time. Our results have direct implications to handle labeled graphs by using single numbers efficiently, which is significant to enable the canonical graph encodings in learning and optimization algorithms. Our bijections are the first known in the literature.

Modularity is vital to engineer complex products and machines. We assert that modularity can emerge in the context of desirable structures constrained to life cycle factors; and propose a method to evaluate machine modularity in the context of life cycle optimization. Experiments using explorit, a new and well suited global optimization algorithm, on fve relevant and divergent machine models show that it is possible to obtain tractable modules within the context of life cycle factors.

Asset selection is a challenging task in the complex global financial system, whose nature has highlighted the need to rethink conventional practices. The attractive and non-toxic assets must be kept on the eye so that our financial systems sustain building blocks in our economic systems. This paper presents an asset selection framework using Genetic Network Programming(GNP). GNP handles evolvable graph structures that prevent the size expansion for dynamic and complex environments, which in turn make it suitable for dealing with decision processes effectively under uncertainty such as partially observable Markov decision processes. Simulations using stocks, bonds and currencies from relevant financial markets in USA, Europe and Asia show the competitive advantages of the proposed method against relevant selection strategies in the finance literature.