Rear wings, front wings and floor edges are examples of or more distinctive parts of the car's structure that must bend in response to aeroelasticity when a car is running in an aerodynamic configuration which does not encompass ground-effect regulation. Even at sub-millimetres scales, these geometric deformations will modify the pressure distributions in aerodynamics and therefore a vehicle's stability at supersonic velocities of over 300 km/h. Even though a number of telemetry solutions exist, using discrete strain gauges and single camera vision pipelines can only be used for local measurement or planar measurement, which cannot provide a complete measurement of the state of structural conditions for real-time SHM. Proposed in this paper is a Multi-Stage and Multi-Camera Fusion (MCF) structure for 3D structural deformation estimation that combines synchronized high-speed cameras, stereo correspondence matching, multi-view triangulation and a temporal learning module based on the Transformer architecture. The proposed system surpasses all baseline systems in terms of both reconstruction accuracy (98.9%) and precision of depth measurement (0.8 mm mean reconstruction error) while maintaining a 38 ms latency, meeting the definition of real-time performance on our aeroelastic benchmark for Formula 1.

Keywords : Formula 1, Multi-Camera Fusion, 3D Reconstruction, Structural Deformation, Aeroelasticity, Transformer.

Authors : Tirlingi Tirupathi Rao

Title : 3D Structural Deformation Estimation Using Multi-Camera Fusion in Formula 1

Volume/Issue : 2025;2(1 ( January - March ))

Page No : 25 - 31

In F1 vehicles, Aero-elastic wing flutter is a serious concern on the integrity of the wing structure that can cause catastrophic failure in the vehicle within milliseconds of the onset of flutter, at race speeds above 300 km/h. The current telemetry-approaches based on convolutional neural networks (CNNs) or long short-term memory (LSTM) models show limited predictions lead time and high false-alarm rates under the highly dynamic aero-mechanical conditions they encounter when handling race events. This paper introduces the Wing Flutter Detection (TF-WFD) framework based on Transformer, combining high-speed vision and multiple sets of aero-elastic parameters from inertial measurement units (IMUs) and fibre Bragg grating (FBG) strain gauge data.This paper proposes a framework of Wing Flutter Detection (TF-WFD) based on Transformer, which integrates heterogeneous multi-sensor telemetry data, including a high-speed vision system, IMUs, and fibre Bragg grating (FBG) strain meter. A real-time risk metric based on the structural deformation, acceleration, and velocity features is derived from a composite Flutter Index (FI), which is interpretable. Understanding how well the framework can perform, it achieves an accuracy of 97.9% in classification, an F1-score of 0.974 in detection, a detection lead time of 180ms (twice that of CNN baselines) and 0.05 false alarms per hour, which allows for proactive aerodynamic intervention before structural damage is imminent.

Keywords : Formula 1 Aerodynamics; Wing Flutter; Transformer; Aero-elasticity; Multi-Head Self-Attention; Sensors.

Authors : J V G Prakasa Rao Pyla

Title : Transformer-Based Real-Time Wing Flutter Detection in Formula 1 Vehicles

Volume/Issue : 2025;2(1 ( January - March ))

Page No : 18 - 24

Although significant advances have been made in classification of electroencephalograms (EEGs) in biomedical signal analysis, supervised approaches require high numbers of labeled data which are expensive and challenging to collect in medical environments. Self-supervised learning has become an exciting alternative, as it can be used to obtain useful representation learning from recordings that are not labeled. In this paper, a self-supervised geometric learning framework is proposed, which combines the Symmetric Positive Definite (SPD) manifold representation and contrastive learning objective for robust EEG classification. Multi-channel EEG-based covariance structures are projected onto Riemannian manifolds, where the nonlinear relationships between neural signals are preserved. The objective, which is a combination of the geometric and self-supervised objectives, significantly decreases the dependency on annotation by 65% and achieves a classification accuracy of 98.2%, outperforming CNN, Transformer and fully supervised baselines. The proposed framework shows high generalizability and is applicable to motor imagery, mental arithmetic and clinical brain-computer interface (BCI) tasks.

Keywords : EEG Classification, Self-Supervised Learning, Geometric Learning, Riemannian Geometry, SPD Manifold.

Authors : Uppe Nanaji , Yam Krishna Poudel

Title : Self-Supervised Geometric Learning for EEG Classification

Volume/Issue : 2025;2(1 ( January - March ))

Page No : 12 - 17

The classification methods based on EEG have been shown to be valuable for EEG diagnosis of neurological diseases and brain-computer interface (BCI) development. However, too often these types of approaches to deeper learning and machine learning for quantum computing are not interpretable and pose challenges for clinical data integration. For medical practitioners, there needs to be clear, clinically useful, explanations before a diagnostic prediction is made automatically. For Sintering a Quantum EEG classification pipeline with an Explainable Artificial Intelligence (XAI) could be beneficial, as shown in this paper. The proposed framework attempts to make the framework simultaneously predictable, while treating it as an interpretable one by multi-band signal preprocessing data, the variational quantum feature would be embedded in the attention-guiding classification and the post-hoc of the modules would be explained. The system in particular utilizes Shapley Additive explanations (SHAP) feature explanation together with multi-head attention, quantum kernel functions to deliver mathematically reliable and clinically sensible explanations. The results of standard test experiments with EEG data, indicate that the network proposed in the work, achieves a type accuracy of 98,4% and an interpretability score of 0,93, which is significantly higher than the classification accuracy of the traditional CNN network, the Transformer network and the Quantum SVM network. The results demonstrate that the proposed system can bring the quantum advantage closer to being explainable and more trusted clinical Artificial Intelligence.

Keywords : Explainable Artificial Intelligence, EEG, QML, SHAP, LIME, Quantum Kernels, BCI, Neural Signal Analysis.

Authors : S S D Maha Lakshmi , Mohd Akbar

Title : Explainable AI Framework for Quantum-Enhanced EEG Signal Classification

Volume/Issue : 2025;2(1 ( January - March ))

Page No : 6 - 11

In this paper, we have come to know that the epilepsy is a common neural disorder in children which should be diagnosed in the early stage and effectively treated. Traditional systems for automatic seizure detection in the context of paediatric electroencephalography (EEG) are complicated by the fact that the EEG signals are non-stationery and artifact prone. Recent deep learning methods have tried to solve this by representing EEG signals as 2D topographic images fed into Convolutional Neural Nets (CNNs), which have a Euclidean-based representation. The assumption ignores intrinsic non-linear spatial correlations of brain activity. The present paper introduces a new quantum classical hybrid model where the multi-channel EEG signals are viewed as trajectories on a non-Euclidean Symmetric Positive Definite (SPD) manifold. We measure the true information-theoretic distance between normal and pre-ictal brain states by computing the Affine-Invariant Riemannian Metric (AIRM) and projecting these geometric properties on a tangent space. These high-dimensional tangent vectors are classified with a Quantum Support Vector Machine (QSVM) that uses amplitude embedding. Experimental results show that the manifold-based geometric model is much more effective than classical CNN-SVM and hybrid CNN-QSVM architectures, and that it is very robust in the presence of noisy environments typical of paediatric EEG recordings.

Keywords : Pediatrics Epilepsy, EEG, SPD, AIRM, QSVM, Geodesic distance

Authors : T Mural Mohan , Mesfin Dagne

Title : Information-Geometric Manifold Learning for Pediatric Epilepsy Detection via Quantum-Enhanced Geodesic Classification

Volume/Issue : 2025;2(1 ( January - March ))

Page No : 1 - 5