Search

Autonomous robots depend crucially on their ability to perceive and process information from dynamic, ever-changing environments. Traditional simultaneous localization and mapping (SLAM) approaches struggle to maintain consistent scene representations because of numerous moving objects, often treating dynamic elements as outliers rather than explicitly modeling them in the scene representation. In this paper, we present a novel hierarchical 3D scene graph-based SLAM framework that addresses the challenge of modeling and estimating the pose of dynamic objects and agents. We use fiducial markers to detect dynamic entities and to extract their attributes while improving keyframe selection and implementing new capabilities for dynamic entity mapping. We maintain a hierarchical representation where dynamic objects are registered in the SLAM graph and are constrained with robot keyframes and the floor level of the building with our novel entity-keyframe constraints and intra-entity constraints. By combining semantic and geometric constraints between dynamic entities and the environment, our system jointly optimizes the SLAM graph to estimate the pose of the robot and various dynamic agents and objects while maintaining an accurate map. Experimental evaluation demonstrates that our approach achieves a 27.57% reduction in pose estimation error compared to traditional methods and enables higher-level reasoning about scene dynamics., Comment: 8 pages, 3 figures, 3 tables

We construct a Neural Network that approximates the matrix multiplication operator for any activation functions for which there exist a Neural Network which can approximate the scalar multiplication function. In particular, we use the Strassen algorithm for reducing the number of weights and layers needed for such Neural Network. This allows us to define another Neural Network for approximating the inverse matrix operator. Also, by relying on the Galerkin method, we apply those Neural Networks for resolving parametric elliptic PDEs for a whole set of parameters at the same time. Finally, we discuss the improvements with respect to prior results.

In this Master Thesis, we study the approximation capabilities of Neural Networks in the context of numerical resolution of elliptic PDEs and Approximation Theory. First of all, in Chapter 1, we introduce the mathematical definition of Neural Networks and perform some basic estimates on their composition and parallelization. Then, we implement in Chapter 2 the Galerkin method using Neural Network. In particular, we manage to build a Neural Network that approximates the inverse of positive-definite symmetric matrices, which allows to get a Garlerkin numerical solution of elliptic PDEs. Finally, in Chapter 3, we introduce the approximation space of Neural Networks, a space which consists of functions in $L^p$ that are approximated at a certain rate when increasing the number of weights of Neural Networks. We find the relation of this space with the Besov space: the smoother a function is, the faster it can be approximated with Neural Networks when increasing the number of weights.

Understanding the relationships between geometric structures and semantic concepts is crucial for building accurate models of complex environments. In indoors, certain spatial constraints, such as the relative positioning of planes, remain consistent despite variations in layout. This paper explores how these invariant relationships can be captured in a graph SLAM framework by representing high-level concepts like rooms and walls, linking them to geometric elements like planes through an optimizable factor graph. Several efforts have tackled this issue with add-hoc solutions for each concept generation and with manually-defined factors. This paper proposes a novel method for metric-semantic factor graph generation which includes defining a semantic scene graph, integrating geometric information, and learning the interconnecting factors, all based on Graph Neural Networks (GNNs). An edge classification network (G-GNN) sorts the edges between planes into same room, same wall or none types. The resulting relations are clustered, generating a room or wall for each cluster. A second family of networks (F-GNN) infers the geometrical origin of the new nodes. The definition of the factors employs the same F-GNN used for the metric attribute of the generated nodes. Furthermore, share the new factor graph with the S-Graphs+ algorithm, extending its graph expressiveness and scene representation with the ultimate goal of improving the SLAM performance. The complexity of the environments is increased to N-plane rooms by training the networks on L-shaped rooms. The framework is evaluated in synthetic and simulated scenarios as no real datasets of the required complex layouts are available., Comment: Submitted to ICRA 2025

The two-dimensional electron gas (2DEG) found in KTaO3-based interfaces has garnered attention due to its remarkable electronic properties. In this study, we investigated the conducting system embedded at the Si3N4/Al//KTO(110) heterostructure. We demonstrate that the Al/KTO interface supports a conducting system, with the Si3N4 passivation layer acting as a barrier to oxygen diffusion, enabling ex-situ characterization. Our findings reveal that the mobility and carrier density of the system can be tuned by varying the Al layer thickness. Using scanning transmission electron microscopy, electron energy-loss spectroscopy, X-ray photoemission spectroscopy, and time-of-flight secondary ion mass spectrometry, we characterized the structural and chemical composition of the interface. We found that the Al layer fully oxidizes into AlOx, drawing oxygen from the KTaO3 substrate. The oxygen depletion zone extends 3-5 nm into the substrate and correlates to the Al thickness. Heterostructures with thicker Al layers exhibit higher carrier densities but lower mobilities, likely due to interactions with the oxygen vacancies that act as scattering centers. These findings highlight the importance of considering the effect and extent of the oxygen depletion zone when designing and modeling two-dimensional electron systems in complex oxides.

Robots navigating indoor environments often have access to architectural plans, which can serve as prior knowledge to enhance their localization and mapping capabilities. While some SLAM algorithms leverage these plans for global localization in real-world environments, they typically overlook a critical challenge: the "as-planned" architectural designs frequently deviate from the "as-built" real-world environments. To address this gap, we present a novel algorithm that tightly couples LIDAR-based simultaneous localization and mapping with architectural plans under the presence of deviations. Our method utilizes a multi-layered semantic representation to not only localize the robot, but also to estimate global alignment and structural deviations between "as-planned" and as-built environments in real-time. To validate our approach, we performed experiments in simulated and real datasets demonstrating robustness to structural deviations up to 35 cm and 15 degrees. On average, our method achieves 43% less localization error than baselines in simulated environments, while in real environments, the as-built 3D maps show 7% lower average alignment error

A promising branch of neuromorphic computing aims to perform cognitive operations in hardware leveraging the physics of efficient and well-established nano-devices. In this work, we present a reconfigurable classifier based on a network of electrically connected magnetic tunnel junctions that categorizes information encoded in the amplitude of input currents through the spin torque driven magnetization switching output configuration. The network can be trained to classify new data by adjusting additional programming currents applied selectively to the junctions. We experimentally demonstrate that a network composed of three magnetic tunnel junctions can learn to classify spoken vowels with a recognition rate that surpasses the performance of software multilayer neural networks with the same number of trained parameters in this task. These results, obtained with the same nano-devices and working principle employed in industrial spin-transfer torque magnetic random-access memories (STT-MRAM), constitute an important step towards the development of large-scale neuromorphic networks based on well-established technology., Comment: To be published in Physical Review Applied. 13 pages, 4 figures

Nonrelativistic Quantum Chromodynamics (NRQCD) breaks down in the region of low transverse momentum, where the transverse momentum of the produced quarkonium state is sensitive to multiple scattering with the incoming hadron and to soft gluon radiation. In this kinematic regime, the transverse-momentum-dependent (TMD) factorization framework is required, promoting the long-distance matrix elements (LDMEs) of NRQCD to the so-called TMD shape functions (TMDShFs), which encode both the soft gluon radiation and the formation of the heavy-quark bound state. In this work, we apply an effective-field theory approach (combining NRQCD and SCET) to the photon-gluon fusion process in inclusive $J/\psi$ leptoproduction. We derive a factorization theorem for the cross section in terms of TMDShFs, compute these quantities at next-to-leading order, establish their evolution, and study their matching onto the corresponding LDMEs in the high-transverse-momentum region. Our results are particularly relevant to the Electron-Ion Collider, where $J/\psi$ leptoproduction can be used to probe gluon transverse-momentum-dependent parton distribution functions (gluon TMDPDFs)., Comment: v2: typos fixed, accepted for publication in JHEP; v1: 48 pages, 7 figures

This paper explores the application of Human-in-the-Loop (HITL) strategies in training machine learning models in the medical domain. In this case a doctor-in-the-loop approach is proposed to leverage human expertise in dealing with large and complex data. Specifically, the paper deals with the integration of genomic data and Whole Slide Imaging (WSI) analysis of breast cancer. Three different tasks were developed: segmentation of histopathological images, classification of this images regarding the genomic subtype of the cancer and, finally, interpretation of the machine learning results. The involvement of a pathologist helped us to develop a better segmentation model and to enhance the explainatory capabilities of the models, but the classification results were suboptimal, highlighting the limitations of this approach: despite involving human experts, complex domains can still pose challenges, and a HITL approach may not always be effective.

A key challenge in realizing fault-tolerant quantum computers is circuit optimization. Focusing on the most expensive gates in fault-tolerant quantum computation (namely, the T gates), we address the problem of T-count optimization, i.e., minimizing the number of T gates that are needed to implement a given circuit. To achieve this, we develop AlphaTensor-Quantum, a method based on deep reinforcement learning that exploits the relationship between optimizing T-count and tensor decomposition. Unlike existing methods for T-count optimization, AlphaTensor-Quantum can incorporate domain-specific knowledge about quantum computation and leverage gadgets, which significantly reduces the T-count of the optimized circuits. AlphaTensor-Quantum outperforms the existing methods for T-count optimization on a set of arithmetic benchmarks (even when compared without making use of gadgets). Remarkably, it discovers an efficient algorithm akin to Karatsuba's method for multiplication in finite fields. AlphaTensor-Quantum also finds the best human-designed solutions for relevant arithmetic computations used in Shor's algorithm and for quantum chemistry simulation, thus demonstrating it can save hundreds of hours of research by optimizing relevant quantum circuits in a fully automated way., Comment: 25 pages main paper + 19 pages appendix

A non-contact linear slider based on stable superconducting magnetic levitation with a long permanent magnet as a slider and two fixed superconducting disks which define the slide way has been designed, built and tested. The slider can be moved stably along a stroke of 11.5mm by supplying a low current in a coil located at the end of the stroke while the levitation remains stable providing a reliable mechanism for linear displacement in a cryogenic environment. The response is linear with a sensitivity of 52 {\mu}m/mA for displacements lower than 6 mm. Run out, pitch, yaw and roll have been measured demonstrating an overall good performance. In all cases the measured hysteresis was lower than 250 {\mu}m and the measured run out was also lower than 250 {\mu}m both for Y and Z axis. Roll and yaw were always below 300 {\mu}rad, that is one order of magnitude lower than the pitch (4500 {\mu}rad).

Pedestrian intention prediction is crucial for autonomous driving. In particular, knowing if pedestrians are going to cross in front of the ego-vehicle is core to performing safe and comfortable maneuvers. Creating accurate and fast models that predict such intentions from sequential images is challenging. A factor contributing to this is the lack of datasets with diverse crossing and non-crossing (C/NC) scenarios. We address this scarceness by introducing a framework, named ARCANE, which allows programmatically generating synthetic datasets consisting of C/NC video clip samples. As an example, we use ARCANE to generate a large and diverse dataset named PedSynth. We will show how PedSynth complements widely used real-world datasets such as JAAD and PIE, so enabling more accurate models for C/NC prediction. Considering the onboard deployment of C/NC prediction models, we also propose a deep model named PedGNN, which is fast and has a very low memory footprint. PedGNN is based on a GNN-GRU architecture that takes a sequence of pedestrian skeletons as input to predict crossing intentions.

The time evolution of a wave packet is a tool to detect topological phase transitions in two-dimensional Dirac materials, such as graphene and silicene. Here we extend the analysis to HgTe/CdTe quantum wells and study the evolution of their electron current wave packet, using 2D effective Dirac Hamiltonians and different layer thicknesses. We show that the two different periodicities that appear in this temporal evolution reach a minimum near the critical thickness, where the system goes from normal to inverted regime. Moreover, the maximum of the electron current amplitude changes with the layer thickness, identifying that current maxima reach their higher value at the critical thickness. Thus, we can characterize the topological phase transitions in terms of the periodicity and amplitude of the electron currents., Comment: 10 pages, 10 figures, 1 table

We report the dynamic characterization of the spin-torque-driven in-plane precession modes of a spin-torque nano-oscillator based on two different synthetic ferrimagnets: a pinned one characterized by a strong RKKY interaction which is exchange coupled to an antiferromagnetic layer; and a second one, non-pinned characterized by weak RKKY coupling. The microwave properties associated with the steady-state precession of both SyFs are characterized by high spectral purity and power spectral density. However, frequency dispersion diagrams of the damped and spin transfer torque modes reveal drastically different dynamical behavior and microwave emission properties in both SyFs. In particular, the weak coupling between the magnetic layers of the non-pinned SyF raises discontinuous dispersion diagrams suggesting a strong influence of mode crossing. An interpretation of the different dynamical features observed in the damped and spin torque modes of both SyF systems was obtained by solving simultaneously, in a macrospin approach, a linearized version of the Landau-Lifshitz-Gilbert equation including the spin transfer torque term., Comment: 16 pages, 8 figures

This work studies a central extremal graph theory problem inspired by a 1975 conjecture of Erd\H{o}s, which aims to find graphs with a given size (number of nodes) that maximize the number of edges without having 3- or 4-cycles. We formulate this problem as a sequential decision-making problem and compare AlphaZero, a neural network-guided tree search, with tabu search, a heuristic local search method. Using either method, by introducing a curriculum -- jump-starting the search for larger graphs using good graphs found at smaller sizes -- we improve the state-of-the-art lower bounds for several sizes. We also propose a flexible graph-generation environment and a permutation-invariant network architecture for learning to search in the space of graphs., Comment: To appear in the proceedings of IJCAI 2024. First three authors contributed equally, last two authors made equal senior contribution

Recent works on SLAM extend their pose graphs with higher-level semantic concepts like Rooms exploiting relationships between them, to provide, not only a richer representation of the situation/environment but also to improve the accuracy of its estimation. Concretely, our previous work, Situational Graphs (S-Graphs+), a pioneer in jointly leveraging semantic relationships in the factor optimization process, relies on semantic entities such as Planes and Rooms, whose relationship is mathematically defined. Nevertheless, there is no unique approach to finding all the hidden patterns in lower-level factor-graphs that correspond to high-level concepts of different natures. It is currently tackled with ad-hoc algorithms, which limits its graph expressiveness. To overcome this limitation, in this work, we propose an algorithm based on Graph Neural Networks for learning high-level semantic-relational concepts that can be inferred from the low-level factor graph. Given a set of mapped Planes our algorithm is capable of inferring Room entities relating to the Planes. Additionally, to demonstrate the versatility of our method, our algorithm can infer an additional semantic-relational concept, i.e. Wall, and its relationship with its Planes. We validate our method in both simulated and real datasets demonstrating improved performance over two baseline approaches. Furthermore, we integrate our method into the S-Graphs+ algorithm providing improved pose and map accuracy compared to the baseline while further enhancing the scene representation.

We compute the gluon transverse-momentum-dependent fragmentation function (TMDFF) at next-to-leading order (NLO) into heavy quarkonium in the color-octet $^3S_1^{[8]}$ channel, based on the NRQCD factorization approach. The spurious rapidity divergences are explicitly shown to cancel in a well-defined TMDFF, which incorporates the needed soft factor. We also compute the integrated gluon FF at NLO in the same $^3S_1^{[8]}$ channel, and show that the matching coefficient of the TMDFF onto the FF at large transverse momentum is the expected one. These results are relevant to perform precise and sensible phenomenological studies of transverse-momentum spectra of quarkonium production, for which the production mechanism through fragmentation plays a relevant role, like in the future Electron-Ion Collider., Comment: v3: matches published version; v2: typos corrected; v1: 29 pages, 2 figures

We analyze the magneto-optical conductivity (and related magnitudes like transmittance and Faraday rotation of the irradiated polarized light) of some elemental two-dimensional Dirac materials of group IV (graphene analogues, buckled honeycomb lattices, like silicene, germanene, stannane, etc.), group V (phosphorene), and zincblende heterostructures (like HgTe/CdTe quantum wells) near the Dirac and gamma points, under out-of-plane magnetic and electric fields, to characterize topological-band insulator phase transitions and their critical points. We provide plots of the Faraday angle and transmittance as a function of the polarized light frequency, for different external electric and magnetic fields, chemical potential, HgTe layer thickness and temperature, to tune the material magneto-optical properties. We have shown that absortance/transmittance acquires extremal values at the critical point, where the Faraday angle changes sign, thus providing fine markers of the topological phase transition. In the case of non-topological materials as phosphorene, a minimum of the transmittance is also observed due to the energy gap closing by an external electric field., Comment: 16 pages, 11 figures, including supplemental material with 7 pages and 6 figures, 3 animated gifs. A new paragraph has been included at the Introduction (the fifth one) to establish the new contributions. We have also added a sentence at the end of the abstract for this purpose. The review of previous results has been extended with a new paragragh in the Introduction (the third one)

In this paper, we propose a solution for graph-based global robot simultaneous localization and mapping (SLAM) using architectural plans. Before the start of the robot operation, the previously available architectural plan of the building is converted into our proposed architectural graph (A-Graph). When the robot starts its operation, it uses its onboard LIDAR and odometry to carry out an online SLAM relying on our situational graph (S-Graph), which includes both, a representation of the environment with multiple levels of abstractions, such as walls or rooms, and their relationships, as well as the robot poses with their associated keyframes. Our novel graph-to-graph matching method is used to relate the aforementioned S-Graph and A-Graph, which are aligned and merged, resulting in our novel informed Situational Graph (iS-Graph). Our iS-Graph not only provides graph-based global robot localization, but it extends the graph-based SLAM capabilities of the S-Graph by incorporating into it the prior knowledge of the environment existing in the architectural plan, Comment: 4 Page Workshop paper for ICRA 2023. arXiv admin note: substantial text overlap with arXiv:2303.02076