Iliana Loi

Simulating fatigue effects on human motion is essential both for conducting biomechanical analyses in order to develop overexertion and fatigue prevention and mitigation techniques, as well as to realistically render temporally evolved fatigued movements in animation and Extended Reality (XR) pipelines. In this work, we propose a framework to account for the dynamic injection of fatigue and fatigue‐induced characteristic features in non‐fatigued movements for data‐driven fatigue‐driven motion synthesis, an under‐explored scientific field in recent literature. To do so, we leverage an interplay between a siamese neural network and a Transformer‐based Fatigue Model to account for the encoding and sampling of fatigue features, while fatigue scalings are incorporated into motion via the state‐of‐the‐art Fatigue‐PINN . Our quantitative evaluation findings confirm the effectiveness and validity of our framework, while qualitative analysis shows that the fatigued motion sequences produced from our model are comparable with the observations of real‐world experimental studies investigating the impact of externally perceived fatigue in human motion. Moreover, we developed a demonstrator to showcase and assess the capability of our model to dynamically integrate fatigue scalings and fatigue features into motion sequences, while evaluating its ability to be seamlessly integrated into XR and motion synthesis environments.

Toxic speech on online civic engagement platforms (CEPs) disproportionately affects marginalized groups and threatens the diversity of citizen voices. However, the deployment of AI-driven toxic speech detection (TSD) tools for CEPs faces complex challenges from legal, psychological, and technical perspectives that remain insufficiently explored. We present a first-of-its-kind interdisciplinary review of these challenges, focusing on the explainability of TSD systems, their compliance with European legal standards and offer a roadmap for ethical deployment. Our review reveals three main findings. First, although transparency in AI decision-making is necessary from both legal and psychological perspectives, assessing the explainability of AI-driven TSD tools, and their compliance with legal regulations within Europe, remains a significant challenge. Second, current explainability approaches, ranging from toxic span identification to advanced explainable AI methods, lack standardized metrics. This makes it difficult to assess their reliability and appropriateness for CEPs. Third, despite the importance of TSD, frameworks and best practices for CEPs are still lacking in existing literature. This paper aims to fill this gap by providing a holistic perspective on the challenges and solutions for TSD deployment. It provides the foundation for collaborative efforts to develop and standardize metrics, evaluation protocols, and best practices that can ensure AI decisions in CEPs are transparent, accountable, and aligned with users’ needs.

Fatigue modeling is essential for motion synthesis tasks to model human motions under fatigued conditions and biomechanical engineering applications, such as investigating the variations in movement patterns and posture due to fatigue, defining injury risk mitigation and prevention strategies, formulating fatigue minimization schemes and creating improved ergonomic designs. Nevertheless, employing data-driven methods for synthesizing the impact of fatigue on motion, receives little to no attention in the literature. In this work, we present Fatigue-PINN, a deep learning framework based on Physics-Informed Neural Networks, for modeling fatigued human movements, while providing joint-specific fatigue configurations for adaptation and mitigation of motion artifacts on a joint level, resulting in more smooth, hence physically-plausible animations. To account for muscle fatigue, we simulate the fatigue-induced fluctuations in the maximum exerted joint torques by leveraging a PINN adaptation of the Three-Compartment Controller model to exploit physics-domain knowledge for improving accuracy. This model also introduces parametric motion alignment with respect to joint-specific fatigue, hence avoiding sharp frame transitions. Our results indicate that Fatigue-PINN accurately simulates the effects of externally perceived fatigue on open-type human movements being consistent with findings from real-world experimental fatigue studies. Since fatigue is incorporated in torque space, Fatigue-PINN provides an end-to-end encoder-decoder-like architecture, to ensure transforming joint angles to joint torques and vice-versa, thus, being compatible with motion synthesis frameworks operating on joint angles.

Fifth-Generation Networks deal with dynamic fluctuations in user traffic and the demands of each connected user and application. This creates a need for optimizing resource allocation to reduce network congestion in densely populated urban centers, and further ensure Quality of Service in 5G environments. To address this issue, we present a framework for both predicting user traffic and allocating users to base stations in 5G networks using neural network architectures. This framework consists of a hybrid approach utilizing a Long Short-Term Memory(LSTM) network or a Transformer architecture for user traffic prediction in base stations, as well as a Convolutional Neural Network (CNN) to allocate users to base stations in a realistic scenario. The models show high accuracy in the tasks performed; especially, in the user traffic prediction task, where the models show an accuracy of over 99%. Overall, our framework is capable of capturing long-term temporal features and spatial features from 5G user data, taking a significant step towards a holistic approach in data-driven resource allocation and traffic prediction in 5G networks.

Artificial intelligence, machine learning, and deep learning as a service have become the status quo for many industries, leading to the widespread deployment of models that handle sensitive data. Well-performing models, the industry seeks, usually rely on a large volume of training data. However, the use of such data raises serious privacy concerns due to the potential risks of leaks of highly sensitive information. One prominent threat is the Membership Inference Attack, where adversaries attempt to deduce whether a specific data point was used in a model's training process. An adversary's ability to determine an individual's presence represents a significant privacy threat, especially when related to a group of users sharing sensitive information. Hence, well-designed privacy-preserving machine learning solutions are critically needed in the industry. In this work, we compare the effectiveness of L2 regularization and differential privacy in mitigating Membership Inference Attack risks. Even though regularization techniques like L2 regularization are commonly employed to reduce overfitting, a condition that enhances the effectiveness of Membership Inference Attacks, their impact on mitigating these attacks has not been systematically explored.

Most recent musculoskeletal dynamics estimation methods are designed for predefined actions, such as gait, and don’t generalize to various tasks. In this work, we address the problem of estimating internal biomechanical forces during more than one actions by introducing unsupervised domain adaptation into a deep learning model. More specifically, we developed a Bidirectional Long Short-Term Memory network for knee contact force prediction, enhanced with correlation alignment layers, in order to minimize the domain shift between kinematic data from different actions. Furthermore, we used the novel Neural State Machine (NSM) as a simulation platform to test and visualize our model predictions in a wide range of trajectories adapted to different 3D scene geometries in real-time. We conducted multiple experiments, including comparison with previous models, model alignment across action classes and real-to-synthetic data alignment. The results showed that the proposed deep learning architecture with domain adaptation performs better than the benchmark in terms of NRMSE and t-test. Overall, our method is capable of predicting knee contact forces for more than one action classes using a single architecture and thereby opens the path for estimating internal forces for intermediate actions, while the knowledge of the hidden state of motion may be used to support personalized rehabilitation. Moreover, our model can be easily integrated into any human motion simulation environment, which shows its potential in enabling biomechanical analysis in an automated and computationally efficient way.

The inference of 3D motion and dynamics of the human musculoskeletal system has traditionally been solved using physics-based methods that exploit physical parameters to provide realistic simulations. Yet, such methods suffer from computational complexity and reduced stability, hindering their use in computer graphics applications that require real-time performance. With the recent explosion of data capture (mocap, video) machine learning (ML) has started to become popular as it is able to create surrogate models harnessing the huge amount of data stemming from various sources, minimizing computational time (instead of resource usage), and most importantly, approximate real-time solutions. The main purpose of this paper is to provide a review and classification of the most recent works regarding motion prediction, motion synthesis as well as musculoskeletal dynamics estimation problems using ML techniques, in order to offer sufficient insight into the state-of-the-art and draw new research directions. While the study of motion may appear distinct to musculoskeletal dynamics, these application domains provide jointly the link for more natural computer graphics character animation, since ML-based musculoskeletal dynamics estimation enables modeling of more long-term, temporally evolving, ergonomic effects, while offering automated and fast solutions. Overall, our review offers an in-depth presentation and classification of ML applications in human motion analysis, unlike previous survey articles focusing on specific aspects of motion prediction.

This study presents a semi-automatic framework to create subject-specific total knee replacement finite element models, which can be used to analyze locomotion patterns and evaluate knee dynamics. In recent years, much scientific attention was attracted to pre-clinical optimization of customized total knee replacement operations through computational modeling to minimize post-operational adverse effects. However, the time-consuming and laborious process of developing a subject-specific finite element model poses an obstacle to the latter. One of this work's main goals is to automate the finite element model development process, which speeds up the proposed framework and makes it viable for practical applications. This pipeline's reliability was ratified by developing and validating a subject-specific total knee replacement model based on the 6th SimTK Grand Challenge data set. The model was validated by analyzing contact pressures on the tibial insert in relation to the patient's gait and analysis of tibial contact forces, which were found to be in accordance with the ones provided by the Grand Challenge data set. Subsequently, a sensitivity analysis was carried out to assess the influence of modeling choices on tibial insert's contact pressures and determine possible uncertainties on the models produced by the framework. Parameters, such as the position of ligament origin points, ligament stiffness, reference strain, and implant-bone alignment were used for the sensitivity study. Notably, it was found that changes in the alignment of the femoral component in reference to the knee bones significantly affect the load distribution at the tibiofemoral joint, with an increase of 206.48% to be observed at contact pressures during 5° internal rotation. Overall, the models produced by this pipeline can be further used to optimize and personalize surgery by evaluating the best surgical parameters in a simulated manner before the actual surgery.