Search

This work presents the experiments and solution outline for our teams winning submission in the Learn To Race Autonomous Racing Virtual Challenge 2022 hosted by AIcrowd. The objective of the Learn-to-Race competition is to push the boundary of autonomous technology, with a focus on achieving the safety benefits of autonomous driving. In the description the competition is framed as a reinforcement learning (RL) challenge. We focused our initial efforts on implementation of Soft Actor Critic (SAC) variants. Our goal was to learn non-trivial control of the race car exclusively from visual and geometric features, directly mapping pixels to control actions. We made suitable modifications to the default reward policy aiming to promote smooth steering and acceleration control. The framework for the competition provided real time simulation, meaning a single episode (learning experience) is measured in minutes. Instead of pursuing parallelisation of episodes we opted to explore a more traditional approach in which the visual perception was processed (via learned operators) and fed into rule-based controllers. Such a system, while not as academically "attractive" as a pixels-to-actions approach, results in a system that requires less training, is more explainable, generalises better and is easily tuned and ultimately out-performed all other agents in the competition by a large margin.

In this study, we propose a robust set of features derived from a thorough research of contemporary practices in voice pathology detection. The feature set is based on the combination of acoustic handcrafted features. Additionally, we introduce pitch difference as a novel feature. We combine this feature set, containing data from the publicly available Saarbr\"ucken Voice Database (SVD), with preprocessing using the K-Means Synthetic Minority Over-Sampling Technique algorithm to address class imbalance. Moreover, we applied multiple ML models as binary classifiers. We utilized support vector machine, k-nearest neighbors, naive Bayes, decision tree, random forest and AdaBoost classifiers. To determine the best classification approach, we performed grid search on feasible hyperparameters of respective classifiers and subsections of features. Our approach has achieved the state-of-the-art performance, measured by unweighted average recall in voice pathology detection on SVD database. We intentionally omit accuracy as it is highly biased metric in case of unbalanced data compared to aforementioned metrics. The results are further enhanced by eliminating the potential overestimation of the results with repeated stratified cross-validation. This advancement demonstrates significant potential for the clinical deployment of ML methods, offering a valuable tool for an objective examination of voice pathologies. To support our claims, we provide a publicly available GitHub repository with DOI 10.5281/zenodo.13771573. Finally, we provide REFORMS checklist., Comment: 33 pages, 8 figures, code repository: https://github.com/aailab-uct/Automated-Robust-and-Reproducible-Voice-Pathology-Detection

We study the quantum thermal transport of elastic excitations through a two-dimensional elastic waveguide between two thermal reservoirs. We solve the classical Kirchhoff-Love equation for rectangular wires and explore the dispersion relation for both the symmetric and antisymmetric solutions. Then, we study the phonon transport of these modes within the second quantization framework by analyzing the mean quadratic displacement, from which the energy density current, the temperature field, and conductance are determined. We identify the relevant modes contributing to thermal transport and explore the average temperature difference to reach the high-temperature limit. We expect our results to pave the way for understanding phonon-mediated thermal transport in two-dimensional mesoscopic quantum devices., Comment: Submitted to Journal of Applied Physics

Aerosols' impact on the performance of a clear-sky solar irradiance model is often evaluated from the perspective of the overall accuracy of estimates. This study assesses the aerosol role in clear-sky solar irradiance modelling from a totally different perspective, namely the ability of a model to accurately separate global solar irradiance into its fundamental direct and diffuse components. In an innovative approach, the analysis is focused on the sensitivity of parametric solar irradiance models to changes in aerosol conditions. A new measure, the aerosol influence quantifier, is introduced for assessing the correlation strength between the relative variation in clear-sky diffuse fraction and the relative variation in atmospheric aerosol load. The effectiveness of the aerosol influence quantifier is explored as a tool for assessing the sensitivity of three parametric clear-sky models to variations in input aerosol data. A rather surprising result is reported, e.g. if high quality aerosol data are not available, one could potentially maximize the estimate's accuracy by using a parametric clear-sky solar irradiance model with low sensitivity to aerosol variations. In such cases, it is shown that the aerosol influence quantifier can be used as a tool for identifying an adequate model for input data with a given uncertainty.

Mapping is crucial for spatial reasoning, planning and robot navigation. Existing approaches range from metric, which require precise geometry-based optimization, to purely topological, where image-as-node based graphs lack explicit object-level reasoning and interconnectivity. In this paper, we propose a novel topological representation of an environment based on "image segments", which are semantically meaningful and open-vocabulary queryable, conferring several advantages over previous works based on pixel-level features. Unlike 3D scene graphs, we create a purely topological graph with segments as nodes, where edges are formed by a) associating segment-level descriptors between pairs of consecutive images and b) connecting neighboring segments within an image using their pixel centroids. This unveils a "continuous sense of a place", defined by inter-image persistence of segments along with their intra-image neighbours. It further enables us to represent and update segment-level descriptors through neighborhood aggregation using graph convolution layers, which improves robot localization based on segment-level retrieval. Using real-world data, we show how our proposed map representation can be used to i) generate navigation plans in the form of "hops over segments" and ii) search for target objects using natural language queries describing spatial relations of objects. Furthermore, we quantitatively analyze data association at the segment level, which underpins inter-image connectivity during mapping and segment-level localization when revisiting the same place. Finally, we show preliminary trials on segment-level `hopping' based zero-shot real-world navigation. Project page with supplementary details: oravus.github.io/RoboHop/, Comment: Published at ICRA 2024; 9 pages, 8 figures

Facial nerve paresis is a severe complication that arises post-head and neck surgery; This results in articulation problems, facial asymmetry, and severe problems in non-verbal communication. To overcome the side effects of post-surgery facial paralysis, rehabilitation requires which last for several weeks. This paper discusses an unsupervised approach to rehabilitating patients who have temporary facial paralysis due to damage in mimetic muscles. The work aims to make the rehabilitation process objective compared to the current subjective approach, such as House-Brackmann (HB) scale. Also, the approach will assist clinicians by reducing their workload in assessing the improvement during rehabilitation. This paper focuses on the clustering approach to monitor the rehabilitation process. We compare the results obtained from different clustering algorithms on various forms of the same data set, namely dynamic form, data expressed as functional data using B-spline basis expansion, and by finding the functional principal components of the functional data. The study contains data set of 85 distinct patients with 120 measurements obtained using a Kinect stereo-vision camera. The method distinguish effectively between patients with the least and greatest degree of facial paralysis, however patients with adjacent degrees of paralysis provide some challenges. In addition, we compared the cluster results to the HB scale outputs.

Strong interaction energy shifts and widths in kaonic atoms are calculated for the first time using microscopic $K^- N$ + $K^- NN$ potentials derived from $K^- N$ scattering amplitudes constructed within SU(3) chiral coupled-channels models of meson-baryon interactions. The in-medium modifications of the free-space amplitudes due to the Pauli correlations are taken into account. The $K^-N+K^-NN$ potentials evaluated for 23 nuclear species are confronted with kaonic atoms data. The description of the data significantly improves when the $K^-NN$ absorption is included. To get $\chi^2$ as low as for the $K^-N+$phenomenological multi-nucleon potential an additional phenomenological term, accounting for $K^{-}-3N(4N)$ processes, is still needed. However, density dependence of this phenomenological term points out some deficiencies in the microscopic potentials and further improvements of the applied model are thus desirable. The calculated branching ratios for $K^-N$ and $K^-NN$ absorption channels in the $^{12}$C$+K^-$ atom are in reasonable agreement with the old bubble chamber data, as well as with the latest data from the AMADEUS Collaboration., Comment: 24 pages, 7 figures, 7 tables

In haptic object discrimination, the effect of gripper embodiment, action parameters, and sensory channels has not been systematically studied. We used two anthropomorphic hands and two 2-finger grippers to grasp two sets of deformable objects. On the object classification task, we found: (i) among classifiers, SVM on sensory features and LSTM on raw time series performed best across all grippers; (ii) faster compression speeds degraded performance; (iii) generalization to different grasping configurations was limited; transfer to different compression speeds worked well for the Barrett Hand only. Visualization of the feature spaces using PCA showed that the gripper morphology and the action parameters were the main source of variance, rendering generalization across embodiment or grasp configurations very hard. On the highly challenging dataset consisting of polyurethane foams alone, only the Barrett Hand achieved excellent performance. Tactile sensors can thus provide a key advantage even if recognition is based on stiffness rather than shape. The dataset with 24000 measurements is publicly available., Comment: 12 pages, 9 figures

Quantum walks exhibit properties without classical analogues. One of those is the phenomenon of asymptotic trapping -- there can be non-zero probability of the quantum walker being localised in a finite part of the underlying graph indefinitely even though locally all directions of movement are assigned non-zero amplitudes at each step. We study quantum walks with the flip-flop shift operator and the Grover coin, where this effect has been identified previously. For the version of the walk further modified by a random dynamical disruption of the graph (percolated quantum walks) we provide a recipe for the construction of a complete basis of the subspace of trapped states allowing to determine the asymptotic probability of trapping for arbitrary finite connected simple graphs, thus significantly generalizing the previously known result restricted to planar 3-regular graphs. We show how the position of the source and sink together with the graph geometry and its modifications affect the excitation transport. This gives us a deep insight into processes where elongation or addition of dead-end subgraphs may surprisingly result in enhanced transport and we design graphs exhibiting this pronounced behavior. In some cases this even provides closed-form formulas for the asymptotic transport probability in dependence on some structure parameters of the graphs., Comment: 15 pages, 12 figures

This study investigates the influence of several Monte Carlo radiation transport codes and nuclear models on the simulation of secondary neutron spectra and its impact on calculating and measuring neutron doses in proton therapy. Three different multi-purpose Monte Carlo radiation transport codes (FLUKA, MCNPX, Geant4) were used with different available nuclear models to calculate secondary neutron energy spectra at various points inside a water tank phantom with PMMA walls using a 10 cm x 10 cm rectangular, mono-energetic proton beam (110, 150, 180, and 210 MeV). Using Kerma approximation secondary neutron doses were calculated applying fluence-to-dose equivalent conversion coefficients in water. Moreover, the impact of varying spectra for electrochemically etched CR39 detector calibration was analyzed for different codes and models. In distal positions beyond the Bragg peak, results show largest variations between the codes, up to 53% for the high energy neutron fluence at 16 cm from the Bragg peak of the 110 MeV proton beam. In lateral positions, the variation between the codes is smaller and for the total neutron fluence within 20%. Variation in the nuclear models in MCNPX was only visible for the proton beam energies of 180 and 210 MeV and modeling the high energy neutron fluence which reached up to 23% for 210 MeV at 11 cm lateral from the beam axis. Impact on neutron dose equivalent was limited for the different models used (<8%) while it was pronounced for the different codes (45% at 16 cm from the Bragg peak of the 110 MeV proton beam). CR39 calibration factors in lateral positions were on average varying 10% between codes and 5 % between nuclear models.This study demonstrated a large impact on the neutron fluence spectra calculated by different codes while the impact of different models in MCNPX proved to be less prominent for the neutron modeling in proton therapy., Comment: 30 pages, 4 figures, 8 tables, accepted for publication in Radiation Measurements