Elektroteknik (E2) // Electrical Engineering (E2)

Använd denna länk för att länka till samlingen:

https://odr.chalmers.se/handle/20.500.12380/12

Arbetar med hållbara och smarta lösningar på samhällsutmaningar, såsom energieffektivitet och elektrifiering inom områden från transport- och produktionssystem till kommunikationslösningar och medicinteknik.

Läs mer om våra utbildningar

För forskning och forskningspublikationer, se https://research.chalmers.se/organisation/elektroteknik/

At the department of Electrical Engineering research and education are performed in the areas of Communication and Antenna systems, Systems and Control, Signal processing and Biomedical engineering, and Electric Power Engineering.

We work with challenges for a sustainable future in society of today, for example in the growing demands concerning efficient systems for communications and electrification. Our knowledge is of use everywhere where there is advanced technology with integrated electronics, no matter if it involves electricity, electrical signals, optical signals or microwaves.

Studying at the Department of Electrical Engineering at Chalmers

For research and research output, please visit https://research.chalmers.se/en/organization/electrical-engineering/

Browse

Visar 1 - 20 av 21

3D Pose Estimation of Football Players
Osterman, Joakim; Sjögren, Olof; Chalmers tekniska högskola / Institutionen för elektroteknik; Svensson, Lennart; Sjöberg, Anders
Abstract In the context of football analytics, video recordings of matches play a crucial role in post-game analysis. However, videos are inherently limited because they only allow viewers to follow the match from the camera’s perspective. This thesis is part of a larger project aimed at creating 3D representations of football matches from video, thus enabling users to view the game from anywhere inside the virtual 3D environment. The larger project consists of three parts. This thesis focuses on estimating the camera parameters, as well as the 3D poses and locations of the players in the video. The other two projects focus on player tracking and player texture generation. A pipeline consisting of camera calibration and pose estimation is proposed, taking video recordings and bounding box annotations as input and predicting camera pa rameters as well as the players’ 3D poses and locations. For camera calibration, a model specifically tailored for cameras viewing football fields is used. The results indicate accurately predicted positions and viewing angles for the estimated camera. Pose estimation is performed using a pre-trained model and results in visually ac curate projections, although perspective ambiguities are present when the 3D poses are viewed from different angles. The main approach for positioning players was to detect when players touched the ground and interpolate the positions for ambigu ous frames. The results are promising, but noise in the depth estimations occurs due to perspective ambiguities. Subsequently, an optional optimization of poses and positions using multi-view triangulation is also presented, showing possibilities for further refinement to ensure realistic and consistent human poses. Future work on pose and location optimization could yield a pseudo-truth dataset for further enhancements to improve overall poses and positions from strictly monocular video.
3D Shape Generation through Point Transformer Diffusion
(2023) Lan, Ji; Chalmers tekniska högskola / Institutionen för elektroteknik; Kahl, Fredrik
Diffusion models are a novel class of generative models, which have demonstrated promising results in 3D point cloud generation. Meanwhile, Transformer-based models have achieved impressive outcomes across various benchmarks in the field of 3D point cloud understanding. However, the integration of a Transformer based on the local self-attention mechanism with a diffusion model for 3D point cloud generation remains unexplored. In this work, we propose Point Transformer Diffusion (PTD), a probabilistic and flexible generative model. PTD integrates the standard Denoising Diffusion Probabilistic Model, adapted for 3D data, with Point Transformer, a local self-attention network specifically designed for 3D point clouds. To enhance PTD’s performance, several adjustments and techniques are implemented to align the Point Transformer model more effectively with the diffusion model and with the specific generation task. Experiments demonstrate PTD’s ability to generate realistic and diverse shapes. Furthermore, the evaluation results of PTD are comparable to, and in some cases even marginally superior to, those achieved by Point-Voxel Diffusion, which is a state-of-the-art approach. We hope that our work will inspire future investigations into architectures that combine diffusion models and Transformers, along with their application to a wide range of 3D generation tasks.
AI for capacity estimation in overhead transmission lines - A new model proposal for deploying AI to do dynamic line rating
(2024) Tydén, Ludvig; Nordin, Jonas; Chalmers tekniska högskola / Institutionen för elektroteknik; Le, Anh Tuan; Balouji, Ebrahim
Abstract Dynamic line rating (DLR) for overhead transmission lines, which estimates the current-carrying capacity based on environmental and operational conditions, presents an opportunity to enhance the efficiency and reliability of electrical grids. This thesis explores the feasibility and application of artificial intelligence (AI) to improve the accuracy and scalability of DLR systems, while also lowering the cost of installation. By leveraging machine learning models, particularly physics-informed neural networks (PINNs), this research aims to lay the foundation of the development of an advanced DLR solution capable of real-time and forecast estimation of the capacity of a transmission line. The initial phase of the thesis focuses on implementing a weather-based model, based on the IEEE-738 standard, to estimate line ratings based on weather parameters such as temperature, wind speed, wind angle and solar radiation, but also parameters such as the electrical current and conductor specific metrics. This model serves as both a practical tool for immediate deployment and a foundational step towards more complex AI models. Following the development of the weather-based model, the research transitions to the integration neural networks, with a perspective of utilizing physics informed neural networks. These models combine the data-driven capabilities of traditional machine learning with the robustness of physical laws governing power transmission. The objective is to enhance the precision and reliability of the DLR system, accommodating non-linear relationships and interactions within the data. The thesis proposes a new model of a DLR system based around the weather model and machine learning. The system consists of several modules that each serve a purpose, data collection, ML model, ML training, weather model and real-time capacity estimation. The findings demonstrate that AI-based DLR systems is a new, interesting approach that can significantly improve the operational efficiency of electrical grids by providing more accurate and adaptive line ratings. This research contributes to the field of power systems by offering a scalable and innovative approach to dynamic line rating, supporting the transition towards a smarter infrastructure. The project has gained interest from key stakeholders such as EON, Vattenfall, and Svenska kraftnät, and has been formed in close contact with their respective specialists in DLR systems. The involvement of these industry leaders underscores the practical relevance and potential impact of this type of research in this domain.
An Auto-Annotation Pipeline for Automotive Data Sequences
(2023) Allgurén, Amanda; Jansfelt, Albin; Chalmers tekniska högskola / Institutionen för elektroteknik; Svensson, Lennart
Abstract The field of autonomous driving advances through the expanding utilization of deep learning methods. Training deep learning models on automotive data sequences allows for predictions of an object’s location and its future movements. Harnessing the benefits of deep learning generally requires accompanying annotations, and since the annotation process poses a significant bottleneck, new methods to mitigate this challenge are urgently needed. To address this issue, we propose an auto-annotation pipeline consisting of three modules. First, a 3D object detector is trained on annotated single-frame data and thereafter applied to each frame in a sequence. Second, a model-based tracker connects the bounding boxes across frames and improves low-confidence detections via filtering. Third, we introduce a smoothing network that further refines the detections by also incorporating future frames. The smoothing network considers both bounding boxes and point clouds. With our smoothing network, we show an improvement in center point, size and rotational error. This progress aligns with the efforts of previous work that have developed pipelines integrating object detection, multi-object tracking, and bounding box refinements. However, we distinguish ourselves by working offline with limited sequence annotations. In particular, our pipeline works with sequences where only one frame is annotated. Additionally, we contribute by proposing random window slides as a data augmentation approach. Our work serves as a baseline for object detection, multi-object tracking and smoothing for the Zenseact Open Dataset.
Attention-based Time Series Forecasting with Limited Data
(2024) Vadström, Gustav; Chalmers tekniska högskola / Institutionen för elektroteknik; Monti, Paolo; Banar, Jafar
Electricity outages are common in electrical power systems, and often caused by natural phenomena, human intervention, or faults in electrical components, such as transformers. A small number of these faults can be predicted by analysing the stream of voltage and current. Forecasting faults in electrical power systems can prevent electricity outages that cause production downtime and capital losses. However, data collected in power systems are usually limited and unbalanced because of the very few historical predictable faults. This study focused on evaluating more recently popular attention-based machine learning models for time series prediction in electrical power systems, in a context where data is a significant limitation. The data was real and consisted of disturbances recorded from power systems over sev eral years, along with documented faults. Two different model architectures were evaluated and compared: the Long short-term memory (LSTM) and the transformer. Three different model instances were trained: using features manually extracted from each disturbance recording, using manually extracted features with pre-training on a similar dataset, and using a signal embedding pipeline attached to each model processing raw waveforms. The results from all six training instances showed that the transformer performed better than the LSTM in terms of evaluation metrics, although the LSTM outputs were more interpretable, because the transformer had higher confidence in its outputs even during false predictions. A bottleneck was found in the small sequence lengths, with improvement shown when utilizing pre training on a similar dataset containing longer sequences. The integrated waveform feature embedding also showed improvement over the manually extracted features.
Comparison of Machine Learning Techniques for Beam Management in 5G New Radio (NR)
(2023) Lundberg, Axel; Svensson, Simon; Chalmers tekniska högskola / Institutionen för elektroteknik; Durisi, Giuseppe; Hallinger, Bengt; Tabeshnezhad, Azadeh
Abstract Aligning beams in the initial access of beam management is a challenging and timeconsuming process. Especially, when the number of antenna elements grow large to compensate for high path loss of millimeter waves. Machine learning methods have successfully been applied to the problem of beam selection and perform much better than traditional methods like exhaustive search. In this thesis, some different machine learning approaches are investigated: decision tree, random forest, Adaptive Boosting (AdaBoost), Support Vector Machine (SVM), Multi-level Perceptron (MLP), Q-learning, Deep Q-Network (DQN) and Double Deep Q-Network (DDQN). Each model is adapted to specific scenarios with different preprocessing steps. A total of three scenarios are explored which have been defined by the 3rd Generation Partnership Project (3GPP): Urban Micro (UMi), Urban Macro (UMa) and Rural Macro (RMa). The UMi and UMa scenarios are both implemented with an explicit city layout containing static receivers. The RMa scenario is uniformly distributed and divided into two datasets: one for static receivers and one for dynamic receivers along tracks. Each scenario has been generated by the stochastic channel model called QuaDRiGa. The aim of the thesis is to provide a fair comparison of machine learning models by testing them on data from one simulator. Results show that random forest and AdaBoost perform best overall on all datasets with up to 90% accuracy when predicting the optimal beam pair, which suggests that the search space can be significantly reduced.
Cross-modal image feature matching between infrared and visual images. Adapting intra-modal feature matching models for cross-modal matching
(2024) Räjert, Tommy; Chalmers tekniska högskola / Institutionen för elektroteknik; Zach, Christopher; Lochman, Yaroslava; Ringdahl, Viktor
Abstract Image feature matching is an essential part to various computer vision applications. Many modern solutions apply machine learning techniques to achieve state-of-theart results. A lesser studied problem is matching image features between images of different modalities. This thesis investigates this problem for the visual–LWIR (long-wave infrared) case by utilizing the matching capabilities of the pre-trained intra-modal models SuperPoint and SuperGlue. This is done by adding interfacing models and additional layers to mitigate problems such as catastrophic forgetting and data biasing in the pre-trained models. These techniques prove only marginally successful compared to the pre-trained models themselves. For training these models, a method for sparse pseudo ground truth point correspondence is proposed, and evaluation is done via pose estimation. This thesis provides insight into some specific methods of transfer learning for the SuperPoint and SuperGlue models, methods for ground truth estimation, and discusses the difficulties faced in this problem. Further studying of this problem may be able to construct improved models for LWIR–visual matching, which would enable more reliable methods for cross-modal camera calibration & registration, localization, and image retrieval, with numerous applications in the automotive, defense, and healthcare industries.
Decentralized Training of 3D Lane Detection with Automatic Labeling Using HD Maps
(2022) Xiao, Zhuqi; Mao, Yadong; Chalmers tekniska högskola / Institutionen för elektroteknik; Zach, Christopher
Explainable machine learning for state-of-health prediction using electric vehicle histogram data
(2022) Carlsson, Eric; Hellqvist, Oskar; Chalmers tekniska högskola / Institutionen för elektroteknik; Wik, Torsten
Exploring Open World Object Detection on Autonomous Driving Image Data
(2024) Olsson, Hanna; Johansson, Lukas; Chalmers tekniska högskola / Institutionen för elektroteknik; Kahl, Fredrik; Henriksson, Jens
Abstract Open world object detection (OWOD) enhances traditional object detection by not only recognizing classes it was trained on but also identifying novel classes as ’unknown’, while also incrementally learning these new classes. Since OWOD was introduced in 2021, various methods have been developed, typically trained and evaluated using benchmark datasets like MS-COCO. In this thesis, we examine the performance of one of the stateof- the-art OWOD methods, PROB, in a new context by applying it to the autonomous driving dataset, Zenseact Open Dataset (ZOD), and explore various strategies to enhance its performance. To evaluate the performance, we apply a standard framework in OWOD, looking at wilderness impact (WI), absolute open set error (A-OSE) and unknown recall (U-recall) for the unknown classes and mean average precision (mAP) for the known classes. Our results demonstrate that PROB exhibits inferior performance across all metrics on ZOD compared to benchmark datasets. Modifications to the initial method revealed that tuning the objectness temperature was unnecessary, while adjusting the class distributions for more even representation improved performance for less common classes. The most significant performance improvement was observed when incorporating curriculum learning, which involves changing the training structure by starting with easier training examples and gradually progressing to more difficult ones. However, neither of these improved methods reach the performance of PROB when applied to benchmark datasets, which can primarily be attributed to ZOD being a very different and challenging dataset. These findings underscore the difficulty of applying OWOD methods to diverse real-world datasets and highlight the need for further research to develop more robust and adaptable detection models.
Football Analysis in VR - Texture Estimation with Differentiable Rendering and Diffusion Models
(2024) Anjou, Filip; Ekström, Albin; Chalmers tekniska högskola / Institutionen för elektroteknik; Svensson, Lennart
Machine Learning for Brain Activity Analysis
(2022) Sundqvist, Elias; Hall, David; Chalmers tekniska högskola / Institutionen för elektroteknik; Monti, Paolo
MTL vs STL: NIR Video metadata classification using self-supervised semi-supervised learning
(2023) Fåhraeus, Gustav; Thörnblom, Adam; Chalmers tekniska högskola / Institutionen för elektroteknik; Hammarstrand, Lars
Abstract The combination of self-supervision and semi-supervision has emerged as a popular research topic in recent years. However, existing studies primarily focus on single-task models trained on datasets where individual images are labeled with a single class, overlooking the challenges associated with multi-class scenarios. In this thesis, we propose a modified S4L framework, which is a self-supervised semi supervised learning framework specifically designed to handle partially labeled data. The modified S4L framework addresses some limitations of previous approaches and demonstrates its effectiveness in both multi-task learning and single-task learning settings. The research focuses on the classification of visual attributes in human subjects, specifically in near-infrared (NIR) images.
Physics-Informed Neural Networks: Solving & Discovering Charge Dynamics in Gaseous High Voltage Insulation - Exploring the use of PINNs for Forward and Inverse Problems within Charge Dynamics in Air Insulation
(2023) Björnson , Carl-Johan; Ågren, Felix; Chalmers tekniska högskola / Institutionen för elektroteknik; Serdyuk, Yuriy; Häger, Christian; Hammarström, Thomas; Hjortstam, Olof
Abstract The development of efficient high-voltage equipment is imperative for minimizing greenhouse gas emissions and saving costs within the energy system. Effective insulation plays a pivotal role in such development and requires an understanding of the performance of gaseous insulators, such as air, under high-voltage stress. Electric discharges and charge transport in gases are modeled using systems of partially differentiable equations (PDEs) and their solutions are traditionally approximated numerically with discretizing methods such as the finite element method (FEM). However, such methods have significant shortcomings including difficulty handling high-dimensional problems, non-smooth behaviors, and inverse problems with hidden physics. An emerging, mesh-free alternative to numerical methods is physics-informed neural networks (PINNs). PINNs solve PDEs using a neural network with the PDE and associated constraints embedded into the network’s loss function and are easily extended to inverse problems. Initial experiments with PINNs for the forward problems related to electric discharges and charge dynamics have shown promising advantages compared to FEM but failed to model strongly non-uniform functions and coupled equations within the domain. This thesis contributes to this research by showing how a variety of performance-enhancing techniques can address the weaknesses of previous works, improving accuracy and enabling the modeling of steeper gradients and coupled PDEs. Additionally, it demonstrates how PINNs can be used to solve inverse problems related to discharge and charge dynamics, discovering both unknown parameters and distributions.
Predictive Performance and Calibration of Deep Ensembles Spread Over Time
(2023) Bodin, Alexander; Meding, Isak; Chalmers tekniska högskola / Institutionen för elektroteknik; Svensson, Lennart
Rank based annotation system for supervised learning in medical imaging
(2023) Tärnåsen, Hanna; Bergström, Herman; Chalmers tekniska högskola / Institutionen för elektroteknik; Häggström, Ida
Abstract Supervised learning has become a common approach for extracting information from images. To effectively train a model, a large amount of labeled data is required. While some image annotation tasks are objective and well-defined, others require the annotators to make a subjective assessment. The difficulty and subjective nature of these annotation tasks cause the standard rating-based annotation techniques to suffer from inconsistencies between annotators, implying that two different annotators could assign highly differing labels based on their personal biases. This thesis’ overarching goal is to provide an alternate rank-based system for annotating subjective data that could be applied to supervised learning, with the hope of increasing the quality of labels. The target application for this project is the annotation of the degree of bronchial wall thickening seen in CT scans of the lungs in patients with chronic obstructive pulmonary disease (COPD). Four potential implementations are compared, and consistency, as well as resource demands, are evaluated in several parts. These include imitating the annotation process with simulation, user evaluation with arbitrary subjective assessments, and lastly evaluating bronchial wall thickenings with radiologists. After evaluation, it is observed that the implementation showing the most potential is one based on the TrueSkill algorithm, which employs Bayesian inference and assumes that underlying scores are not definite but instead follows a normal distribution. The findings presented in this thesis indicate a clear increase in inter-annotator agreement for this rank-based system and the study demonstrates that the indirect approach of evaluating images creates more reliable labels than the direct ratingbased method
Towards Federated Fleet Learning Leveraging Unannotated Data
(2023) Viala Bellander, Alexander; Ghafir, Yazan; Chalmers tekniska högskola / Institutionen för elektroteknik; Graell I Amat, Alexandre
Abstract The rapid advancement of autonomous driving technology poses new challenges, including the efficient management and use of large volumes of data generated by autonomous vehicles. federated learning, which allows for distributed, on-device learning, has emerged as a potential solution. However, the effectiveness of federated learning in the context of autonomous driving, particularly when faced with scarce or non-existent labelled data, is still an open question. This thesis explores this issue, employing semi-supervised and imitation learning methodologies within the federated learning framework for autonomous driving tasks such as ego-road segmentation and trajectory prediction. This approach deviates from the conventional assumption of abundant labelled data, aiming instead to maximise on-device learning from unlabelled data. While our experiments demonstrate the potential of federated learning in autonomous driving, results indicate that its performance is currently on par with or slightly less effective than traditional methods for the tasks we studied. Furthermore, this research underscores the largely untapped potential of self-supervised learning methodologies within the federated learning framework for autonomous driving. We argue that further exploration in this area could result in significant breakthroughs and contribute to a future where autonomous vehicles can collectively learn without compromising privacy and efficiency.
Tracking players and ball in football videos
(2024) Ganelius, Hugo; Humayun, Jhanzaib; Chalmers tekniska högskola / Institutionen för elektroteknik; Svensson, Lennart; Lennart
Uncertainty estimation in multi-modal 3D object detection
(2024) Rosén, Anton; Chalmers tekniska högskola / Institutionen för elektroteknik; Hammarstrand, Lars; Johnander, Joakim; Fatemi, Maryam; Lindström, Carl
Abstract Object detection is an important part of many autonomous driving systems, providing condensed information about the vehicle’s surroundings. For good performance in different environmental conditions, multi-modal object detection is often used, where information from different sensors are fused. Due to factors such as sensor noise, occlusion, and adverse weather conditions, there is an inherent uncertainty in the object detection task. Most state-of-the-art approaches for multi-modal 3D object detection do not model these uncertainties for regression. Explicitly modeling and estimating the uncertainties leads to higher interpretability, allows analysis of difficult situations, and can improve the performance of downstream tasks. In this work, we explore how uncertainty can be modeled and estimated in multimodal 3D object detection. We show that directly modeling the uncertainties of bounding box parameters can provide meaningful uncertainty estimates without sacrificing neither predictive performance nor computational efficiency. We compare modeling the uncertainties both separately per detection, using normally distributed random vectors, and jointly per frame, using Poisson multi-Bernoulli random finite sets. Our results show that separate modeling enhances predictive performance, while joint modeling yields more accurate uncertainty estimates. Additionally, we demonstrate that these predicted uncertainties can identify unlabeled data where the model performs poorly, underscoring their importance for more interpretable and safe autonomous driving systems.
Vehicle-to-Everything Optimization Considering Battery Degradation
(2022) Bjurek, Kalle; Hagman, Victor; Chalmers tekniska högskola / Institutionen för elektroteknik; Zou, Changfu; Zou, Changfu

Browse

Browsar Elektroteknik (E2) // Electrical Engineering (E2) efter Program "Data science and AI (MPDSC), MSc"

Sökresultat per sida

Sortera efter