Search

This paper explores the integration of incremental curriculum learning (ICL) with deep reinforcement learning (DRL) techniques to facilitate mobile robot navigation through task-based human instruction. By adopting a curriculum that mirrors the progressive complexity encountered in human learning, our approach systematically enhances robots' ability to interpret and execute complex instructions over time. We explore the principles of DRL and its synergy with ICL, demonstrating how this combination not only improves training efficiency but also equips mobile robots with the generalization capability required for navigating through dynamic indoor environments. Empirical results indicate that robots trained with our ICL-enhanced DRL framework outperform those trained without curriculum learning, highlighting the benefits of structured learning progressions in robotic training.

Memristors, which are characterized by their unique input-voltage-history-dependent resistance, have garnered significant attention for the exploration of next-generation in-memory computing, reconfigurable logic circuits, and neural networks. Memristors are controlled by the applied input voltage; however, the latent potential of their magnetic field sensitivity for spintronics applications has rarely been explored. In particular, valuable functionalities are expected to be yielded by combining their history dependence and magnetic field response. Here, for the first time, we reveal a giant memory function based on the magnetic field history of memristive switching, with an extremely large magnetoresistance ratio of up to 32,900% under a constant bias voltage, using a two-terminal Ge-channel device with Fe/MgO electrodes. We attribute this behavior to colossal magnetoresistive switching induced by the d0 ferromagnetism of Mg vacancies in the MgO layers and impact ionization breakdown in the Ge substrate. Our findings may lead to the development of highly sensitive multi-field sensors, high-performance magnetic memory, and advanced neuromorphic devices., Comment: 30 pages, 5 figures in the main text, 8 figures in Supporting Information

Spin Berry curvature characterizes the band topology as the spin counterpart of Berry curvature and is crucial in generating novel spintronics functionalities. By breaking the crystalline inversion symmetry, the spin Berry curvature is expected to be significantly enhanced; this enhancement will increase the intrinsic spin Hall effect in ferromagnetic materials and, thus, the spin-orbit torques (SOTs). However, this intriguing approach has not been applied to devices; generally, the extrinsic spin Hall effect in ferromagnet/heavy-metal bilayer is used for SOT magnetization switching. Here, SOT-induced partial magnetization switching is demonstrated in a single layer of a single-crystalline Weyl oxide SrRuO3 (SRO) with a small current density of ~3.1{\times}10^6 A cm-2. Detailed analysis of the crystal structure in the seemingly perfect periodic lattice of the SRO film reveals barely discernible oxygen octahedral rotations with angles of ~5{\deg} near the interface with a substrate. Tight-binding calculations indicate that a large spin Hall conductivity is induced around small gaps generated at band crossings by the synergy of inherent spin-orbit coupling and band inversion due to the rotations, causing magnetization reversal. Our results indicate that a minute atomic displacement in single-crystal films can induce strong intrinsic SOTs that are useful for spin-orbitronics devices., Comment: 39 pages, 5 figures in the main text, 9 figures in Supporting Information

Interconversion between charge and spin currents is a key phenomenon in realizing next-generation spintronic devices. Highly efficient spin-charge interconversion is expected to occur at band crossing points in materials with large spin-orbit interactions due to enhanced spin Berry curvature. On the other hand, if defects and/or impurities are present, they affect the electronic band structure, which in turn reduces the spin Berry curvature. Although defects and impurities are generally numerous in materials, their influence on the spin Berry curvature and, consequently, spin-charge interconversion has often been overlooked. In this paper, we perform spin-pumping experiments for stoichiometric SrRuO3 and non-stoichiometric SrRu0.7O3 films at 300 K, where the films are in paramagnetic states, to examine how Ru composition deviation from the stoichiometric condition influences the spin-to-charge conversion, showing that SrRuO3 has a larger spin Hall angle than SrRu0.7O3. We derive the band structures of paramagnetic SrRuO3 and SrRu0.75O3 using first-principles calculations, indicating that the spin Hall conductivity originating from the spin Berry curvature decreases when the Ru deficiency is incorporated, which agrees with the experimental results. Our results suggest that point-defect- and impurity control is essential to fully exploit the intrinsic spin Berry curvature and large spin-charge interconversion function of materials. These insights help us with material designs for efficient spin-charge interconversions.

In this work, we present Curled-Dreamer, a novel reinforcement learning algorithm that integrates contrastive learning into the DreamerV3 framework to enhance performance in visual reinforcement learning tasks. By incorporating the contrastive loss from the CURL algorithm and a reconstruction loss from autoencoder, Curled-Dreamer achieves significant improvements in various DeepMind Control Suite tasks. Our extensive experiments demonstrate that Curled-Dreamer consistently outperforms state-of-the-art algorithms, achieving higher mean and median scores across a diverse set of tasks. The results indicate that the proposed approach not only accelerates learning but also enhances the robustness of the learned policies. This work highlights the potential of combining different learning paradigms to achieve superior performance in reinforcement learning applications., Comment: Paper accepted for 24th International Conference on Control, Automation and Systems (ICCAS)

This paper introduces novel deep reinforcement learning (Deep-RL) techniques using parallel distributional actor-critic networks for navigating terrestrial mobile robots. Our approaches use laser range findings, relative distance, and angle to the target to guide the robot. We trained agents in the Gazebo simulator and deployed them in real scenarios. Results show that parallel distributional Deep-RL algorithms enhance decision-making and outperform non-distributional and behavior-based approaches in navigation and spatial generalization., Comment: Paper accepted at the 24th International Conference on Control, Automation and Systems (ICCAS)

Kolmogorov-Arnold Networks (KANs) have shown potential as an alternative to Multi-Layer Perceptrons (MLPs) in neural networks, providing universal function approximation with fewer parameters and reduced memory usage. In this paper, we explore the use of KANs as function approximators within the Proximal Policy Optimization (PPO) algorithm. We evaluate this approach by comparing its performance to the original MLP-based PPO using the DeepMind Control Proprio Robotics benchmark. Our results indicate that the KAN-based reinforcement learning algorithm can achieve comparable performance to its MLP-based counterpart, often with fewer parameters. These findings suggest that KANs may offer a more efficient option for reinforcement learning models., Comment: Paper accepted at 24th International Conference on Control, Automation and Systems (ICCAS)

Recent advancements in real-time object detection frameworks have spurred extensive research into their application in robotic systems. This study provides a comparative analysis of YOLOv5 and YOLOv8 models, challenging the prevailing assumption of the latter's superiority in performance metrics. Contrary to initial expectations, YOLOv5 models demonstrated comparable, and in some cases superior, precision in object detection tasks. Our analysis delves into the underlying factors contributing to these findings, examining aspects such as model architecture complexity, training dataset variances, and real-world applicability. Through rigorous testing and an ablation study, we present a nuanced understanding of each model's capabilities, offering insights into the selection and optimization of object detection frameworks for robotic applications. Implications of this research extend to the design of more efficient and contextually adaptive systems, emphasizing the necessity for a holistic approach to evaluating model performance., Comment: 11th IEEE International Conference on Cybernetics and Intelligent Systems (CIS)

This paper introduces YamaS, a simulator integrating Unity3D Engine with Robotic Operating System for robot navigation research and aims to facilitate the development of both Deep Reinforcement Learning (Deep-RL) and Natural Language Processing (NLP). It supports single and multi-agent configurations with features like procedural environment generation, RGB vision, and dynamic obstacle navigation. Unique to YamaS is its ability to construct single and multi-agent environments, as well as generating agent's behaviour through textual descriptions. The simulator's fidelity is underscored by comparisons with the real-world Yamabiko Beego robot, demonstrating high accuracy in sensor simulations and spatial reasoning. Moreover, YamaS integrates Virtual Reality (VR) to augment Human-Robot Interaction (HRI) studies, providing an immersive platform for developers and researchers. This fusion establishes YamaS as a versatile and valuable tool for the development and testing of autonomous systems, contributing to the fields of robot simulation and AI-driven training methodologies., Comment: Preprint of 33rd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)

Because of the absence of indistinguishability constraint, interparticle interactions between nonidentical particles have in general much more variety than those between identical particles. In particular, it is known that there exists a $U(2)$ family of two-body contact interactions between nonidentical particles in one spatial dimension. This paper studies breakdown of continuous scale invariance to discrete scale invariance under this $U(2)$ family of two-body contact interactions in two-body problems of nonidentical particles on the half line. We show that, in contrast to the corresponding identical-particle problem, there exist two distinct channels that admit geometric sequences of two-body bound states., Comment: 8 pages, 1 eepic figure

Recent studies have shown that the magnetic properties of the ferromagnetic perovskite oxide La$_{2/3}$ Sr$_{1/3}$ MnO$_3$ (LSMO) grown on an SrTiO3 (STO) substrate, such as its magnetic moment and Curie temperature, can be improved by releasing the film from the substrate. However, the microscopic origin of this enhancement is not yet well understood. In this study, we use synchrotron radiation measurements to investigate the mechanism of ferromagnetism enhancement in an LSMO membrane released from an STO substrate by dissolving a water-soluble Sr$_4$Al$_2$O$_7$ buffer layer. Using resonant photoemission spectroscopy on the as-grown LSMO film and LSMO membrane, we elucidate that the strain release from the STO substrate enhances the itineracy of the Mn-3d electrons via p-d hybridization, and this strengthens the double-exchange interaction. The reinforcement of the double-exchange interaction, in turn, improves the ferromagnetism of LSMO.

While many unsupervised learning models focus on one family of tasks, either generative or discriminative, we explore the possibility of a unified representation learner: a model which addresses both families of tasks simultaneously. We identify diffusion models, a state-of-the-art method for generative tasks, as a prime candidate. Such models involve training a U-Net to iteratively predict and remove noise, and the resulting model can synthesize high-fidelity, diverse, novel images. We find that the intermediate feature maps of the U-Net are diverse, discriminative feature representations. We propose a novel attention mechanism for pooling feature maps and further leverage this mechanism as DifFormer, a transformer feature fusion of features from different diffusion U-Net blocks and noise steps. We also develop DifFeed, a novel feedback mechanism tailored to diffusion. We find that diffusion models are better than GANs, and, with our fusion and feedback mechanisms, can compete with state-of-the-art unsupervised image representation learning methods for discriminative tasks - image classification with full and semi-supervision, transfer for fine-grained classification, object detection and segmentation, and semantic segmentation. Our project website (https://mgwillia.github.io/diffssl/) and code (https://github.com/soumik-kanad/diffssl) are available publicly., Comment: Website: see https://mgwillia.github.io/diffssl/ . Code: see https://github.com/soumik-kanad/diffssl . The first two authors contributed equally. 27 pages, 10 figures, 17 tables. Submission under review. (this article supersedes arXiv:2307.08702)

It is known that three-body contact interactions in one-dimensional $n(\geq3)$-body problems of nonidentical particles can be topologically nontrivial: they are all classified by unitary irreducible representations of the pure twin group $PT_{n}$. It was, however, unknown how such interactions are described in the Hamiltonian formalism. In this paper, we study topologically nontrivial three-body contact interactions from the viewpoint of the path integral. Focusing on spinless particles, we construct an $n(n-1)(n-2)/3!$-parameter family of $n$-body Hamiltonians that corresponds to one particular one-dimensional unitary representation of $PT_{n}$. These Hamiltonians are written in terms of background Abelian gauge fields that describe infinitely-thin magnetic fluxes in the $n$-body configuration space., Comment: 12 pages, many tikz figures; typos corrected

Recent observation of quantum transport phenomena of Weyl fermions has brought much attention to 4d ferromagnetic perovskite SrRuO3 as a magnetic Weyl semimetal. Besides, the hump-like Hall effect anomaly, which might have a topological origin, has also been reported for this material. Here, we show that the emergence of such phenomena is governed by the degree of scattering determined by the defect density (Ru-deficiency- and/or interface-driven-defect scattering) and measurement temperature (phonon scattering), where the former is controlled by varying the growth conditions of the SrRuO3 films in molecular beam epitaxy as well as the film thickness. The resulting electronic transport properties can be classified into three categories: clean, intermediate, and dirty regimes. The transport of Weyl fermions emerges in the clean regime, whereas that of topologically trivial conduction electrons in the ferromagnetic metal state prevail in the intermediate and dirty regimes. In the clean and intermediate regimes, anomalous Hall resistivity obeys a scaling law incorporating the intrinsic Karplus-Luttinger (Berry phase) and extrinsic side-jump mechanisms. The hump-like Hall effect anomaly is observed only in the dirty regime, which is contrary to the scaling law between anomalous Hall resistivity and longitudinal resistivity. Hence, we conclude that this anomaly is not inherent to the material and does not have a topological origin. We also provide defect- and temperature-dependent transport phase diagrams of stoichiometric SrRuO3 and Ru-deficient SrRu0.7O3 where the appearance of Weyl fermions and hump-like Hall effect anomaly are mapped. These diagrams may serve as a guideline for designing SrRu1-xO3-based spintronic and topological electronic devices.

Inspired by the covering-space method in path integral on multiply connected spaces, we here present a universal formula of time-evolution kernels for continuous- and discrete-time quantum walks on orbit spaces. In this note, we focus on the case in which walkers' configuration space is the orbit space $\Lambda/\Gamma$, where $\Lambda$ is an arbitrary lattice and $\Gamma$ is a discrete group whose action on $\Lambda$ has no fixed points. We show that the time-evolution kernel on $\Lambda/\Gamma$ can be written as a weighted sum of time-evolution kernels on $\Lambda$, where the summation is over the orbit of initial point in $\Lambda$ and weight factors are given by a one-dimensional unitary representation of $\Gamma$. Focusing on one dimension, we present a number of examples of the formula. We also present universal formulas of resolvent kernels, canonical density matrices, and unitary representations of arbitrary groups in quantum walks on $\Lambda/\Gamma$, all of which are constructed in exactly the same way as for the time-evolution kernel., Comment: 23 pages, 5 eepic figures; typos corrected, discussions improved